Generative Adversarial Networks (GANs) deliver significant perceptual quality improvements over super-resolution methods using simple reconstruction losses. By combining adversarial training with perceptual loss functions, they avoid the over-smoothing resulting from the convergence towards the average of all plausible solutions. However, the likelihood of mode collapse if generator and discriminator learning rates are not carefully tuned or the training data does not represent the entire solution space, often introduces unpleasant artifacts that can compromise the overall quality of the upscaled images. We propose a new GAN architecture trained on the statistical distribution of a residue resulting from subtracting a spatial interpolation of the low-resolution (LR) image from the high-resolution (HR) ground-truth, instead of directly learning the pixel distribution of the HR image conditioned to the LR input. Since the statistical distribution of this residue is easier to model, the probability of mode collapse and consequent introduction of artifacts are mitigated, resulting in an algorithm competitive with much more computationally complex diffusion or transformer models.

In this paper, we propose a new 3D pattern matching based video compression algorithm. Spatiotemporal prediction tools are used to exploit both the temporal and the spatial redundancies, and the resulting residue is encoded using a 3D data coding extension of the Multidimensional Multiscale Parser (MMP) algorithm. MMP was originally proposed as a generic lossy data compression algorithm. A high degree of adaptivity and versatility allowed it to be competitive with state-of-the-art transform-based compression methods for a wide range of applications and data sets. The performance of the proposed algorithm is compared with that of H.264/AVC, achieving close results for some data sets, which indicates the potential of the pattern matching paradigm as an alternative to the traditional hybrid video codecs.

In this paper, we propose a pattern-matching-based algorithm for video compression. This algorithm, named multidimensional multiscale parser (MMP)-Video, is based on the H.264/AVC video encoder, but uses a pattern-matching paradigm instead of the state-of-the-art transform-quantization-entropy encoding approach. The proposed method adopts the use of multiscale recurrent patterns to compress both spatial and temporal prediction residues, totally replacing the use of transforms and quantization. Experimental results show that the coding performance of MMP-Video is better than the one of H.264/AVC high profile, especially for medium to high bit-rates. The gains range up to 0.7 dB, showing that, in spite of its larger computational complexity, the use of multiscale recurrent pattern matching paradigm deserves being investigated as an alternative for video compression.

In this paper, we propose a new encoder for scanned compound documents, based upon a recently introduced coding paradigm called multidimensional multiscale parser (MMP). MMP uses approximate pattern matching, with adaptive multiscale dictionaries that contain concatenations of scaled versions of previously encoded image blocks. These features give MMP the ability to adjust to the input image's characteristics, resulting in high coding efficiencies for a wide range of image types. This versatility makes MMP a good candidate for compound digital document encoding. The proposed algorithm first classifies the image blocks as smooth (texture) and nonsmooth (text and graphics). Smooth and nonsmooth blocks are then compressed using different MMP-based encoders, adapted for encoding either type of blocks. The adaptive use of these two types of encoders resulted in performance gains over the original MMP algorithm, further increasing the performance advantage over the current state-of-the-art image encoders for scanned compound images, without compromising the performance for other image types.

In this paper we present a new segmentation method for the multidimensional multiscale parser (MMP) algorithm. In previous works we have shown that, for text and compound images, MMP has better compression efficiency than state-of-the-art transform-based encoders like JPEG2000 and H.264/AVC; however, it is still inferior to them for smooth images. In this paper we improve the performance of MMP for smooth images by employing a more flexible block segmentation scheme than the one defined in the original algorithm. The new partition scheme allows MMP to exploit the image's structure in a much more adaptive and effective way. Experimental tests have shown consistent performance gains, mainly for smooth images. When employing the new block segmentation scheme, MMP outperforms the state-of-the-art JPEG2000 and H.264/AVC Intra-frame image coding algorithms for both smooth and non-smooth images, at low to medium compression ratios.