<p dir="ltr">In recent decades, the rapid growth of internet traffic, particularly driven by high-definition images/videos has highlighted the critical need for effective image compression to reduce bit rates and enable efficient data transmission. Learned lossy image compression (LIC), which uses end-to-end deep neural networks, has emerged as a highly promising method, even outperforming traditional methods such as the intra-coding of the versatile video coding (VVC) standard. This thesis contributes to the field of LIC in two ways. First, we present a theoretical bound-guided knowledge distillation technique, which utilizes estimated bound information rate-distortion (R-D) functions to guide the training of LIC models. Implemented with a modified hierarchical variational autoencoder (VAE), this method demonstrates superior rate-distortion performance with reduced computational complexity. Next, we introduce a token mixer neural architecture, referred to as <i>contextual clustering</i>, which serves as an alternative to conventional convolutional layers or self-attention mechanisms in transformer architectures. Contextual clustering groups pixels based on their cosine similarity and uses linear layers to aggregate features within each cluster. By integrating with current LIC methods, we not only improve coding performance but also reduce computational load. </p>

This diploma thesis deals with digital image lossy compression. Lossy compression in general inserts some kind of distorsion to the resulting image. The distorsion should not be interupting or even noticable in the better case. For image analysis there is used process called transformation and for choosing relevant coefficients process called coding. Evaluation of image quallity can be done by objective or subjective method. There is encoder introduced and realized in this work. Encoder utilizes two-dimension wavelet transform and SPIHT algortihm for coefficient coding. It was made use of accelerated method of wavelet transform computation by lifting scheme. Coder can proccess color information of images using modificated original SPIHT algorithm. For implementation the JAVA programming language was employed. The object-oriented design principes was made use of and thus the program is easy to extended. At demonstaration pictures there are shown effectiveness and characteristic way of distorsion of the proposed coder at high compression rates.