Texture Synthesis Based Image Coding Algorithm

Image and video based multimedia interaction have become a de facto mode of communication of today's ubiquitous information appliances. Almost all cell phones and laptop computers are now equipped with digital cameras. The amount of digital images and digital video production has grown exponentially, as can be witnessed by the success of YouTube. Underneath all these glamorous development, a key enabling information technology is the JPEG and MPEG families of international image and video coding standards. Examples such as JPEG-2000, MPEG-?, and H.264/AVC have revolutionized how digital image and video are stored, exchanged, and displayed by millions of users over the globe.Despite these great achievements, it is widely accepted that current prevailing image and video algorithms still left much room for further improvement. For example, motion compensation, intra-prediction and transform are employed in current video coding to exploit spatial-temporal redundancy based on the mean squared error (MSE) criterion. But the types of redundancies exploited by current video coding schemes are rather limited since they mainly focus on pixel-wise redundancy rather than perceptual redundancy. It is widely accepted that perceptual quality and pixel-by-pixel fidelity need not refer to the same thing.A brightness-adapted patch-based texture synthesis (BAPTS) algorithm is presented. In the nature images, different regions within the same texture may have varying brightness. Because of the limitation of the sample size, the synthesized texture by traditional patch-based texture synthesis (PTS) methods looks like with constant brightness which degrades the subjective image quality seriously. Distinct from traditional PTS methods, BAPTS features a new brightness adaptation algorithm that retains texture structure and brightness fidelity of the synthesized texture to that of the original image. It is more suitable for image compression. A modified H.264/AVC intra coding algorithm is presented. It employs an L-shaped template based prediction algorithm to predict intra coding model of 4×4 block. Compared to the H.264/AVC standard, it uses the spatial correlation of neighboring blocks more efficiency. The prediction accuracy is improved, and the modified algorithm's coding efficiency is higher than that of H.264/AVC. Extensive experiments have been performed, the average bit rate saving is 5.90%(BD-Rate), the average bit rate saving in lowly compressed case is 2.46%, and the average bit rate saving in highly compressed case is 9.45%.A novel image compression algorithm based on the brightness-adapted patch-based texture synthesis (BAPTS) approach is presented. In this method, highly textured regions of the image are automatically segmented and compressed using BAPTS to achieve comparable perceptual quality at lower bit rate. In order to select the optimal texture patch, it is necessary to compare the quality of the synthesized texture for each candidate texture patches. In this work, a Gabor wavelet transform based similarity (distance) measure is adopted to gauge the quality of the synthesized textured region. The background of the image is compressed by modified H.264/AVC intra coding algorithm. The BAPTS based image coding algorithm uses perceptual redundancy to improve coding efficiency. Compared to H.264/AVC intra coding algorithm, the proposed method achieves 37% bit rate saving when with the same subjective image quality.As the development of Internet and mobile communication technologies, image and video based multimedia interaction are more and more popular in today's mobile networks. Due to different kinks of fading and multipath interference for the wireless channels, mobile video communication experiences burst packet losses. A GOP-level transmission distortion modeling is presented in this work. It uses linear-fitting approach to estimate transmission distortion for different frames in a GOP. Unequal loss protections are applied to important frames. According to extensive experiments, the proposed algorithm is a fast and accurate importance assessment tool in allocating limited channel resources optimally for the mobile streaming video.