Filter design for gray image by down-sample and up-sample

Abstract

ABSTRACT Aims : Designing of a filter which can separate out the target from background in a given image. Methods : The original gray image is down-sampled by rejecting every altern ate pixel values between two columns and two rows. A 3-step down-sampling was done to avoid major information loss. A 3-step up-sampling was done by replicating the lower row and the right column from the down-sampled data matrix to obtain the original size of the matrix. The image matrix thus obtained wa s subtracted from th e original image. Results : The iterative down-sampling and up-sampling matrix gives the background information. Subtraction from original image obtains the target. Thus filters the background. Filter design, Target detection, Up-sample, Down-sample, Intensity discontinuity . 1. INTRODUCTION Given an image, perceiving a target is based on filter design. Once the targets are recognized the other tasks like feature extraction for classification can be done. In this study I have concentrated on designing a filter which can separate out background and object information from a given image. The image processing usually involves segmentation by edge detection taking the peaks of small objects and ignoring the background. I hypothesize that it may possible to homogenize the background. If one assumes that back-ground is one which has the modal population of pixel values then there could be an automated method of doing so then an image matrix could be obtained that is filled predominantly by the modal pixel value. Such a matrix would thus be devoid of the small objects or those whose pixel values occur with less frequency. A subtraction of the main image by the generated image matrix would bring the small objects without the background. I believe that this process has biological equivalence that can recognize both blurred objects and cartoons that accentuate the prominent edges of an object e.g. face [1]. This investigation has been done for detecting the targets in the presence of cluttered environment. In this paper I have tried to design a filter by down-sampling and up-sampling the data for target detection. This kind of approach was first developed by Christopher A. Segall [2] where he used up- sampling and down-sampling of data in the image processing for spatial scalability. Unlike most developed target detection algorithms that require an analysis of the pixel histogram [3]. In such a case the pixel values are arranged in an increasing order disregarding the spatial position in the image matrix. I propose here a very simple and novel down-sampling and up- sampling approach to design a filter for target detection. The proposed method entails creating a sparse matrix using degeneracy of the matrix in multiple cycles and up-sampling by creating new columns by averaging the down-sampled matrix columns and rows. This would homogenize the matrix with background (modal pixel values). Though some patents on up-sampling and down-sampling are available on World Wide Web [4] [5] [6] [7], none seem to be using the principle of homogenization or the method outlined below.