Image Reduction using Edge Based Region of Interest

: Region of Interest in an image reduction is significant problem in DCT based image compression, which introduces the apparent distortion of visual quality. In this paper, an effective region of interest based reduction on an image is proposed. Compared with the existing method, first the edges are identified for compression. Secondly a range of ROI type is introduced according to DCT co-efficient distribution. An adaptive compound method is carried out to reduce the region on the image.


INTRODUCTION
In this paper, a novel and systematic scheme is proposed for reducing the ROI reconstruction in transform domain. Adaptive creation of ROI, deduction of ROI reconstruction is carried out by an adaptive compound method. The adaptive choice of concrete method is based on the value of the criterion, derived according to the DCT co-efficient distributions, which are in half ROI size shifted image block and also in the original target image blocks. Compared with some existing methods], the proposed method keeps better visual effects, no matter under low bit rate encoding scenarios, which was measured with subjective method and also objective method.
The remaining paper is organized as follows. Section II introduces the method of locating the edge in compression scheme. Section III describes performing deduction of ROI reconstruction in DCT domain. Experiment results and conclusion of the paper is given in section IV.

LOCATING EDGES
In this section, a fast and efficient edge detection which is carried out directly with the DCT co-efficient is briefly introduced.
Based on the fourth equation calculations are made and the estimation of the rough edge orientation of the certain ROI. The relationship between S mn and the measurement Table I.
The final edge orientation of the certain block will choose the largest Tθ.  The above computation cost will only be 18 multiplications (M) and 27 additions (A).An exact edge map of the compressed image will be extracted with an expressive speed which will contribute a lot to the later deduction ROI operation.

DEDUCTION OF ROI
Based on the above map information, the whole image is classified into two areas (Edge, No edge).The de region of interest will only be done in the area of No edge. In a common platform the ROI effect is introduced by truncating those undesired high frequency DCT components. If the distributions of the pixel values near the ROI boundary accords with linear functions.
Let firstly consider the horizontal situation by choosing two adjacent blocks as the de region of interest units (a n ,b n ).The right four columns of a n and the left four columns of b n from a new image region of interest c n which can be expressed as C is the half-block size shifted image block, which can need to deal with if any abrupt change happens must deduct region of Internet solutions are following either zero-masking scheme or curve replacing scheme (1) and (2) to reconstruct the low frequency part of the C to be a linear one. However those methods will lose some details or introduce some distortions under higher This will directly utilized to reducing the ROI reduction, where range2 is 0.2 and range3 is 0.4. 3. When CR is under the range: [rangeC3 ≠0] Assume rangeC3=1.0 ROI Co-efficients of C are not really suffered and the correlation is not required for further processing.

RESULTS
The proposed algorithms are performed with number of monochrome images. The performance of the proposed method is illustrated under different bits per pixel based on the resulting subjective image quality and also objective measurement in PSNR.
The results are shown in the table. It can be seen that the proposed method is better than the method under many classic condition with different BPP values. As a result, the proposed algorithm outperforms some classic methods in ROI reduction. The efficiency is verified under both the subjective and objective measurements. It is suitable for fast image region of interest compression.