Objective Image Quality Measures Research Articles

Objective image quality measurement by local spatial‐frequency wavelet analysis

An essential determinant of the value of surrogate digital images is their quality. Image quality measurement has become crucial for most image processing applications. Over the past years, there have been many attempts to develop models or metrics for image quality that incorporate elements of human visual sensitivity. However, there is no current standard and objective definition of spectral image quality. This paper proposes a reliable automatic method for objective image quality measurement by local spatial‐frequency wavelet analysis. The analysis is performed locally by dividing an image into 64×64 pixel blocks, and transforming the data into wavelet domain and sub‐band image domain. A fast lifting wavelet algorithm is developed for computationally efficient spatial‐frequency deconstruction of images to extract the features of edges in sub‐band datasets. Wavelet analysis throughout the spatial‐frequency range with respect to noise, sharpness, brightness, contrast, and modulation transfer function (MTF) reveals more detailed information to relate the quality of an image to the interpretation by using an artificial neural network to optimize weightings and results in a promising quantitative measure of image quality. Experimental results from using this method for image quality measurement exhibit good correlation to subjective visual quality assessments. The proposed approach with fast computation provides a potential metric for reliable image quality quantification, allowing image quality to be automatically evaluated following image capture/compression, but prior to permanent storage in a database.

Evaluation of Image Corrected by Retinex Method Based on S-CIELAB and Gazing Information

The purpose of this research is to propose an effective color metric which can predict the perceptual image quality for Retinex method. In this paper, we first give a brief introduction of three kinds of typical single Retinex methods to improve the color reproduction. And then, we state the process for obtaining the observer rating value from the subjective evaluation experiment performed under the sRGB illumination condition. Next, we introduce the S-CIELAB metric and propose a new metric on the basis of S-CIELAB metric that considers the gazing information. The average S-CIELAB color differences with and without the consideration of gazing information were calculated as the objective image quality measures. The correlations between the observer rating values and the objective image quality measures were calculated. The result shows that all of the average S-CIELAB color differences based on the gazing information are better correlated to the observer rating value than the average S-CIELAB color difference over the whole area. The average S-CIELAB color difference weighted by the gazing frequency over the gazing area shows the strong correlation with the observer rating value.

Retrospective analysis of a detector fault for a full field digital mammography system

This paper describes objective and subjective image quality measurements acquired as part of a routine quality assurance (QA) programme for an amorphous selenium (a-Se) full field digital mammography (FFDM) system between August-04 and February-05. During this period, the FFDM detector developed a fault and was replaced. A retrospective analysis of objective image quality parameters (modulation transfer function (MTF), normalized noise power spectrum (NNPS) and detective quantum efficiency (DQE)) is presented to try and gain a deeper understanding of the detector problem that occurred. These measurements are discussed in conjunction with routine contrast-detail (c-d) results acquired with the CDMAM (Artinis, The Netherlands) test object. There was significant reduction in MTF over this period of time indicating an increase in blurring occurring within the a-Se converter layer. This blurring was not isotropic, being greater in the data line direction (left to right across the detector) than in the gate line direction (chest wall to nipple). The initial value of the 50% MTF point was 6 mm−1; for the faulty detector the 50% MTF points occurred at 3.4 mm−1 and 1.0 mm−1 in the gate line and data line directions, respectively. Prior to NNPS estimation, variance images were formed of the detector flat field images. Spatial distribution of variance was not uniform, suggesting that the physical blurring process was not constant across the detector. This change in variance with image position implied that the stationarity of the noise statistics within the image was limited and that care would be needed when performing objective measurements. The NNPS measurements confirmed the results found for the MTF, with a strong reduction in NNPS as a function of spatial frequency. This reduction was far more severe in the data line direction. A somewhat tentative DQE estimate was made; in the gate line direction there was little change in DQE up to 2.5 mm−1 but at the Nyquist frequency the DQE had fallen to approximately 35% of the original value. There was severe attenuation of DQE in the data line direction, the DQE falling to less than 0.01 above approximately 3.0 mm−1. C-d results showed an increase in threshold contrast of approximately 25% for details less than 0.2 mm in diameter, while no reduction in c-d performance was found at the largest detail diameters (1.0 mm and above). Despite the detector fault, the c-d curve was found to pass the European protocol acceptable c-d curve.

A comparison between objective and subjective image quality measurements for a full field digital mammography system

This paper presents pre-sampling modulation transfer function (MTF), normalized noise power spectrum (NNPS) and detective quantum efficiency (DQE) results for an amorphous selenium (a-Se) full field digital mammography system. MTF was calculated from the image of an angled 0.5 mm thick Cu edge, acquired without additional beam filtration. NNPS data were acquired at detector air-kerma levels ranging from 9.1 µGy to 331 µGy, using a standard mammography x-ray spectrum of 28 kV, Mo/Mo target/filter combination and 4 cm of PMMA additional filtration. Prior to NNPS estimation, the image statistics were assessed using a variance image. This method was able to easily identify a detector artefact and should prove useful in routine quality assurance (QA) measurements. Detector DQE, calculated from the NNPS and MTF data, dropped to 0.3 for low detector air-kerma settings but reached an approximately constant value of 0.6 above 50 µGy at the detector. Subjective image quality data were also obtained at these detector air-kerma settings using the CDMAM contrast-detail (c-d) test object. The c-d data reflected the trend seen in DQE, with threshold contrast increasing at low detector air-kerma values. The c-d data were then compared against predictions made using two established models, the Rose model and a standard signal detection theory model. Using DQE(0), the Rose model gave results within approximately 15% on average for all the detector air-kerma values studied and for detail diameters down to 0.2 mm. Similar agreement was also found between the measured c-d data and the signal detection theory results, which were calculated using an ideal human visual response function and a system magnification of unity. The use of full spatial frequency DQE improved the agreement between the calculated and observer results for detail sizes below 0.13 mm.

Subjective and objective measures of image quality in digital fluoroscopy

There are numerous methods for assessing image quality in diagnostic X ray. In our study we assessed how imaging quality assurance methods perform in practice. Physics assessments were based on IPEM protocols using Leeds test objects. Clinical assessment was based on a questionnaire. A total of 15 systems in three European locations were assessed, covering a range of image intensifier-TV digital fluoroscopy units. Analysis of 274 clinical questionnaires showed that clinical and physics assessments did not place systems in the same order, based on a given image quality parameter. In almost all the comparisons, low level correlation was measured for statistical comparison of rank order (rs < 0.3). However, broad agreement was observed between physics and clinical assessments for image quality associated with contrast and noise. This study emphasises the importance of maintaining links with clinical assessment, when developing quality assurance metrics, and measuring the mutual performance of clinical and physical assessments of image quality.

The DIMOND project and its impact on radiation protection

The DIMOND III project comprised work packages and subprojects to research digital imaging, interventional radiology and interventional cardiology, i.e. areas where there has been rapid technological change, not matched by radiation protection research. Whilst new practices have great potential benefit, there are risks to patients and staff if the technology is implemented inappropriately. DIMOND aimed to develop generic technology assessment and optimisation tools. Clinical quality criteria and technical parameters were included so that digital imaging procedures can be introduced appropriately. Frequency and dimensions of acceptance and constancy testing for digital imaging systems were studied, along with differences between objective measures of image quality and subjective indices. Patient dosimetry was performed to propose and to establish reference values. Clinical evaluation projects included cardiology, interventional radiology and digital mammography. As a result of the DIMOND project, quality criteria for several procedures were developed. Example results are presented in this paper.

Image quality measurements in radiology

Image quality measurement methods are reviewed and difficulties in various approaches are highlighted. The main emphasis of the paper is on objective image quality measurements, however, subjective assessment methods are also discussed briefly.

Color image zooming on the Bayer pattern

A zooming framework suitable for single-sensor digital cameras is introduced and analyzed in this paper. The proposed framework is capable of zooming and enlarging data acquired by single-sensor cameras that employ the Bayer pattern as a color filter array (CFA). The approach allows for operations on noise-free data at the hardware level. Complexity and cost implementation are thus greatly reduced. The proposed zooming framework employs: 1) a spectral model to preserve spectral characteristics of the enlarged CFA image and 2) an adaptive edge-sensing mechanism capable of tracking the underlying structural content of the Bayer data. The framework readably unifies numerous solutions which differ in design characteristics, computational efficiency, and performance. Simulation studies indicate that the new zooming approach produces sharp, visually pleasing outputs and it yields excellent performance, in terms of both subjective and objective image quality measures.

A new CFA interpolation framework

The paper introduces a new color filter array (CFA) interpolation method for digital still cameras. The proposed interpolation scheme is able to (i) overcome the hardware limitations of existing CFA based image acquisition solutions, and (ii) restore color images with excellent visual quality. The scheme employs an adaptive edge-sensing mechanism which operates along the vertical, horizontal and diagonal directions to correctly interpolate unavailable color components. Building on the computed edge-sensing map and a refined color-difference model, a new correlation–correction algorithm is introduced. In addition to the basic model, adaptively determined correction operations are also discussed and analyzed. The solutions proposed here, described in a novel vector notation, constitute a unique CFA interpolation framework, which readily unifies previous, seemingly unrelated, results. Simulation studies indicate that the proposed method is computationally efficient and yields excellent performance, in terms of subjective and objective image quality measures, while outperforming state-of-the-art CFA interpolation methods.

A robust, cost-effective post-processor for enhancing demosaicked camera images

This paper presents an efficient post-processing/enhancement solution capable of reducing visual artifacts introduced during the image demosaicking process. Edge-sensing weights and the original color filter array data are used to detect structural elements in the captured image, and correct color components generated by the demosaicking process using adaptive, spectral model-based enhancement operations. The solution produces excellent results in terms of both objective and subjective image quality measures.

Demosaicked Image Postprocessing Using Local Color Ratios

A postprocessing method for the correction of visual demosaicking artifacts is introduced. The restored, full-color images previously obtained by cost-effective color filter array interpolators are processed to improve their visual quality. Based on a localized color ratio model and the original underlying Bayer pattern structure, the proposed solution impressively removes false colors while maintaining image sharpness. At the same time, it yields excellent improvements in terms of objective image quality measures.

Objective measurements of image quality

Tissue harmonic imaging (THI) and compound imaging have been reported clinically to improve contrast resolution, tissue differentiation and overall image quality. However, there have been limited studies to date to quantify objectively the improvements in image quality achieved with these new imaging techniques. The aim of this study was to quantify differences in image quality that exist between conventional B-mode imaging, harmonic imaging, compound imaging and harmonic compound imaging. An ATL HDI 5000 scanner with three probes (C5-2, L7-4 and L12-5) was tested with two different types of test object, the Gammex-RMI model 404 GS LE and the Gammex-RMI 403 GS LE. The measurement limitations associated with subjective analysis methods were not present in this study because an automated image analysis program was used to determine the image quality parameters. Therefore, subtle differences between the four imaging modes could be detected. Significant improvements in lateral resolution and slice thickness as a function of depth were found with THI. Contrast resolution and anechoic target detection improved with compound imaging, and harmonic compound imaging improved lateral resolution, slice thickness as a function of depth and contrast resolution. (E-mail: jacinta.browne@northglasgow.scot.nhs.uk)

Adaptive video filtering framework

AbstractWe present a new 3D adaptive filtering approach capable of detecting and removing impulsive noise in image/video sequences. The proposed method takes advantage of switching median schemes and robust lower‐upper‐middle (LUM) smoothing characteristics. Simulation studies reported in this article indicate that the proposed filtering scheme achieves an excellent trade‐off between noise attenuation and detail preserving characteristics, and clearly outperforms previously introduced approaches in terms of subjective and objective image quality measures. Besides the filter analysis and the testing of its performance, an important part of this article discusses the filter implementation in Altera field programmable logic devices (FPLD). Simulation studies indicate that the proposed method can be efficiently implemented in hardware and is suitable for real‐time image/video processing applications. © 2005 Wiley Periodicals, Inc. J Imaging Syst Technol 14, 223–237, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ima.20027

Video quality assessment based on structural distortion measurement

Objective image and video quality measures play important roles in a variety of image and video processing applications, such as compression, communication, printing, analysis, registration, restoration, enhancement and watermarking. Most proposed quality assessment approaches in the literature are error sensitivity-based methods. In this paper, we follow a new philosophy in designing image and video quality metrics, which uses structural distortion as an estimate of perceived visual distortion. A computationally efficient approach is developed for full-reference (FR) video quality assessment. The algorithm is tested on the video quality experts group Phase I FR-TV test data set.

Three-dimensional coronary MR angiography performed with subject-specific cardiac acquisition windows and motion-adapted respiratory gating.

In coronary MR angiography, data are conventionally accepted in only short and fixed periods of the cardiac and respiratory cycles. We hypothesized that a more flexible and subject-specific approach to cardiac and respiratory gating may shorten scanning times while maintaining image quality. We implemented an acquisition technique that uses subject-specific acquisition windows in the cardiac cycle and a motion-adapted gating window for respiratory navigator gating. Cardiac acquisition windows and trigger delays were determined individually from a coronary motion scan. Motion-adapted gating used a 2-mm acceptance window for the central 35% of k-space and a 6-mm window for the outer 65% of k-space. In 10 subjects, three-dimensional coronary MR angiograms of the right and left coronary systems were acquired with this technique (the "adaptive technique") as well as a conventional acquisition method, and the scanning times and image quality were compared. The adaptive technique was then applied prospectively to 40 patients who underwent coronary radiographic angiography. Scanning times with the adaptive technique were reduced by a factor of 2.3 for the right coronary artery and by a factor of 2.2 for the left coronary artery system compared with the conventional technique, mainly because we were able to use longer subject-specific acquisition windows in patients with low heart rates. Subjective and objective measurements of image quality showed no significant differences between the two techniques. Prospective evaluation of MR angiograms yielded a sensitivity and specificity of 74.3% and 88.2%, respectively, to detect significant coronary artery stenoses. Coronary MR angiography with subject-specific acquisition windows and motion-adapted respiratory gating reduces scanning times while maintaining image quality and provides high diagnostic accuracy for the detection of coronary artery stenosis.

Colour-difference based demosaicked image postprocessing

An efficient postprocessing method capable of reducing visual artifacts introduced during the colour filter array interpolation process is presented. The method produces excellent results in terms of both objective and subjective image quality measures.

A survey of patient dose and image quality for computed tomographyscanners in Wales

The main aim of the study was to assess the current level of radiation dosefrom computed tomography scanning in the NHS in Wales and to compare theseresults with previous studies in Wales and the UK. In addition, therelationship between patient dose and image quality was investigated bycomparing mean patient dose calculations with image noise using two quality assurance (QA)phantoms. The results show that although the introduction of spiral scannershas reduced the dose per examination, the collective dose per scanner hasactually increased. The results also highlight the potential for relating doseand objective measures of image quality to assist in the selection of scanningparameters to optimise dose without compromising image quality.

Geodesic Paths Approach to Color Image Enhancement

New filter class for multichannel image processing is introduced and analyzed. The new technique of image enhancement is capable of reducing impulsive and Gaussian noise and it significantly outperforms the standard methods of noise reduction. In the paper a smoothing operator, based on a random walk model and a fuzzy similarity measure between pixels connected by a digital geodesic path is introduced. The efficiency of the proposed method was tested on the standard color images using the objective image quality measures. Obtained results show that the new method not only outperforms standard noise reduction algorithms, but has some interesting features useful for segmentation of noisy color images.

Methods and materials for the measurement of subjective and objective measurements of image quality.

The evaluation of medical imaging systems may be performed using both subjective (e.g. high contrast limiting resolution, and low contrast detail detectability), and objective/physical (e.g. characteristic curve, signal-to-noise ratio, modulation transfer function, and Wiener spectra) indices of image quality. For the purposes of Quality Assurance subjective measurements currently have a wider usage in the clinical environment due to the inherent simplicity of the measurements; however, more rigorous objective approaches are gaining acceptability, and represent a paradigm for possible future QA measurements. Both approaches have their advantages and disadvantages. There is currently wide debate as to the interpretation of these measurements. However, there appears to be a lack of standardisation among scientists in the various approaches to making these measurements, and especially so in the case of objective indices; furthermore, the measurement protocol used must have a bearing on the interpretation of subsequent results. The purpose of this paper is to (i) present the most fundamental indices in the characterisation of 'image quality', and (ii) recommend approaches to the measurement of these indices.

Soft-Copy Quality Control of Digital Spot Images Obtained by Using X-ray Image Intensifiers

To evaluate 12 x-ray image intensifier (XRII) digital spot systems. Four objective tests were performed to assess XRII digital spots: entrance exposure, patient exposure, soft-copy gray scale, and pixel noise. Two additional tests were performed to assess high-contrast limiting resolution and threshold contrast detection. Digital spot XRII entrance exposures averaged 1 x 10(-7) C/kg (0.38 mR) for units with large fields of view (FOVs); mean entrance exposure in a medium-sized patient was 1. 25 x 10(-5) C/kg (48 mR). Luminance measurements of the table-side monitors provided a mean of 473 just-noticeable differences in gray scale with the room lights off. Mean resolution with a bar test pattern was measured as 1.5 line pairs per millimeter for systems with a 40-cm FOV. Measured pixel noise (in relative units) was 6-25. Mean threshold contrast with the lights off was 0.85%. Once input exposure is normalized for FOV and image matrix size, soft-copy assessment of limiting resolution with either low-contrast detection or, preferably, an off-line noise metric (pixel SD) provides objective measurements of digital spot image quality. With the lights on, 10 systems with room-light sensors had an 11% loss of gray scale. For systems without sensors, the loss was 33%.