Matched-texture coding for structurally lossless compression

Abstract

We propose a new texture-based compression approach that relies on new texture similarity metrics and is able to exploit texture redundancies for significant compression gains without loss of visual quality, even though there may visible differences with the original image (structurally lossless). Existing techniques rely on point-by-point metrics that cannot account for the stochastic and repetitive nature of textures. The main idea is to encode selected blocks of textures -as well as smooth blocks and blocks containing boundaries between smooth and/or textured regions -by pointing to previously occurring (already encoded) blocks of similar textures, blocks that are not encoded in this way, are encoded by a baseline method, such as JPEG. Experimental results with natural images demonstrate the advantages of the proposed approach.

Perceptual Texture Similarity for Machine Intelligence Applications

“…They developed an algorithm for structurally lossless compression known as Matched-Texture Coding [119]. In this algorithm, a texture patch is copied from another patch of the image, if the similarity score, measured by STSIM, is above a certain threshold.…”

Improved side matching for matched-texture coding

“…The criteria are meansquared-error (MSE), log variance ratio (LVR), and hierarchical implementations of STSIM-2 [5]. The MSE is a natural choice for facilitating blending and has been used before in [1,2]. However, the main issues are to find side matches that have a good chance of success (target match), and to make sure that we do not miss any good matches (and the associated compression gains).…”

“…The key idea of MTC is finding large image blocks that can be replaced with previously encoded blocks that have similar structure. The encoding of such blocks can be done very efficiently without any significant loss in overall image quality, since there is no encoding of a residual, and as will explain below, only a few bits are needed for referencing the previously encoded block [1]. The remaining blocks are encoded [1] have identified two basic versions of MTC, side matching (SM) and direct block matching (DBM).…”

“…The encoding of such blocks can be done very efficiently without any significant loss in overall image quality, since there is no encoding of a residual, and as will explain below, only a few bits are needed for referencing the previously encoded block [1]. The remaining blocks are encoded [1] have identified two basic versions of MTC, side matching (SM) and direct block matching (DBM). The SM approach attempts to identify blocks that can be replaced by previously encoded blocks by matching the upper and left boundary (side) of the target block with the corresponding boundary of the candidate block.…”

“…Matched-Texture coding (MTC) [1,2,3] is a recently proposed technique that attempts to exploit the redundancy of textured regions in natural images in order to achieve structurally lossless compression at low encoding rates. The key idea of MTC is finding large image blocks that can be replaced with previously encoded blocks that have similar structure.…”