Block-Matched And Edge-Directed Image Interpolation

With the continuous development of electronic technology and the continuous improvement of digital information technology, the digital image has been involved to more and more areas of social life. However, due to the restrictions of the conditions and the equipments, we can generally get lower resolution digital image. Therefore, get higher resolution image from the low-resolution image processing has been acquired by more and more applications. As an effective way to improve the image resolution, image interpolation technique has been widely applied in areas such as medical image processing, remote sensing image processing, and consumer electronics image processing and other areas.The existing classical interpolation algorithms are generally based on the idea of the low-pass filter. Such algorithms are simple to be implemented and generally have low complexity. Due to the limitations of the low-pass filter, these algorithms are not good enough to deal with local details such as edges, textures and other image intense transition areas and generally produce fuzzy edges and generate serration. Because eyes are more sensitive to the edge of image, the algorithms are whether good or bad dependent largely on the processed edges of the image, so people gradually proposed the interpolation algorithms based on the edge direction of the image, NEDI is a representative method proposed by Li et al. NEDI is based on two assumptions of local stability property of covariance and geometric duality, and estimated the covariance coefficients via adaptive aggressive model, and estimated high-resolution pixels by low-resolution pixels and the covariance coefficients. Adaptive edge directed interpolation can better avoid blurred edges and prevent jagged edges from generating. However, since during the interpolation process, equally treat all the pixels in search area would introduce mismatched geometrical duality and mismatched local stability property of covariance into the estimation of the unknown pixels, so that the interpolation result will deviate from the actual situation.This paper fully considered the two mismatched problems proposed above, mis-matched local stability property of covariance and mismatched geometric duality. And proposed a new iterative algorithm based on block-matched for images interpolation. Through the algorithm of block-matched, the effect of mismatched blocks in estimation of high resolution pixels can be reduced and excluded. Block-matched algorithms are mainly depending on the redundant information of nature images. As each sub-block can find similar sub-blocks around itself, which can be used to improve the effect of image processing, this has been widely used in image denoising, image restoration and image interpolation and other image processing areas.At the end of this paper, we compare the effect of the BMEDI algorithm via objective and subjective experiment data. We show the PSNR values and SSIM images. Experimental results show that the new algorithm can suppress the artifacts and preserve image edge much better.