Abstract
Fractal image compression (FIC) is a very popular technique in image compression applications due to its simplicity and superior performance. However, it has a major drawback, which is the long encoding time. This is due to the requirement of performing huge similarity search for encoding each small portion in the image. Thus, reducing the search time of FIC while keeping the quality of reconstructed images at acceptable level is still an active research topic. Therefore, this paper has focused on studying the search problem of the conventional full-search FIC algorithm and the impact of employing a spatial dynamic search technique instead with the matching threshold strategy. Unlike the conventional full-search method that is a spatially static where the search starts from a fixed position (normally from the top-left corner of the image to the bottom-right corner) regardless of the position of the range block being encoding, the idea of the dynamic search method is simple, but effective, and it is based on starting the search from the closest domain block to the range block that needs to be encoded. These two search schemes are tested under different matching threshold values, in which the search is terminated whenever a domain block with an acceptable matching level is found. To make the study comprehensive, the test is performed for different image sizes and types, range block and partitioning step sizes, and quantization levels. The experimental results show the significant impact of using the dynamic search method instead of the conventional search method specifically when the threshold is large. For the best encoding parameters, the improvement amount that can be achieved is near to 90 % in terms of search reduction and 1.6 dB PSNR in terms of image quality.
Highlights
Nowadays, images are produced in huge numbers from countless devices used in our daily life, such as mobile phones, surveillance systems, medical devices, etc
As full-search Fractal image compression (FIC) method is usually much simpler and has relatively higher compression rate and image quality compared to the partial- and non-search based methods, this paper aims to study its search problem and the advantage of the dynamic search approach in comparison to the conventional static search approach
The 256 × 256 images are Lena, LiftingBody, CameraMan and LivingRoom, and the 512 × 512 images are Goldhill, Lake, Aerial and Elaine. These images with different sizes and textures are used to make the study more comprehensive. The performance of both search methods is analyzed using (1) the number of searches required for encoding an image, and (2) the peak signal-to-noise ratio (PSNR) for the decoded image
Summary
Images are produced in huge numbers from countless devices used in our daily life, such as mobile phones, surveillance systems, medical devices, etc. The amount of data produced every single minute is extremely huge, making the process of storing these data on digital devices a very costly problem. Transmitting such big data through the internet network requires a high-speed and unlimited data package, which adds an additional cost into the operating expenses. Fractal algorithm is a lossy compression method which is reliant on portioned iteration function system (PIFS) It attempts to find out the similarities between the image blocks and use them for encoding the image [8].
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
