The split-inference paradigm divides an artificial intelligence (AI) model into two parts. This necessitates the transfer of intermediate feature data between the two halves. Here, effective compression of the feature data becomes vital. In this paper, we employ Z-score normalization to efficiently recover the compressed feature data at the decoder side. To examine the efficacy of our method, the proposed method is integrated into the latest Feature Coding for Machines (FCM) codec standard under development by the Moving Picture Experts Group (MPEG). Our method supersedes the existing scaling method used by the current standard under development. It both reduces the overhead bits and improves the end-task accuracy. To further reduce the overhead in certain circumstances, we also propose a simplified method. Experiments show that using our proposed method shows 17.09% reduction in bitrate on average across different tasks and up to 65.69% for object tracking without sacrificing the task accuracy.

As numerous edge devices start implementing intelligent components, the challenges of energy consumption, bandwidth efficiency, and privacy gain significance. One proposed solution relies on the paradigm of split inference, which optimizes the delegation of the computational load between edge and remote devices. We developed and implemented the standard-compliant split inference system with an encoder and decoder capable of real-time streaming and processing. Our system outperforms state-of-the-art video compression implementations by an average of 83% bitrate reduction, while preserving privacy. We demonstrate the system's real-time performance on consumer devices, with interactive visualizations of object detection and segmentation, incorporating real-time metrics. Demo video: https://youtu.be/bmCbUo_ZWWU

Efficient storing and retrieval of medical images has direct impact on reducing costs and improving access in cloud-based health care services. JPEG 2000 is currently the commonly used compression format for medical images shared using the DICOM standard. However, new formats such as high efficiency video coding (HEVC) can provide better compression efficiency compared to JPEG 2000. Furthermore, JPEG 2000 is not suitable for efficiently storing image series and 3-D imagery. Using HEVC, a single format can support all forms of medical images. This paper presents the use of HEVC for diagnostically acceptable medical image compression, focusing on compression efficiency compared to JPEG 2000. Diagnostically acceptable lossy compression and complexity of high bit-depth medical image compression are studied. Based on an established medically acceptable compression range for JPEG 2000, this paper establishes acceptable HEVC compression range for medical imaging applications. Experimental results show that using HEVC can increase the compression performance, compared to JPEG 2000, by over 54%. Along with this, a new method for reducing computational complexity of HEVC encoding for medical images is proposed. Results show that HEVC intra encoding complexity can be reduced by over 55% with negligible increase in file size.

A method is introduced to enhance the High Efficiency Video Coding early skip decision process. The proposed method exploits visual acuity and high-velocity motion perception of the human visual system (HVS). The method determines coding units where the HVS is unlikely to notice the changes and hence can be skipped. In this paper, the focus is mobile devices, such as smart phones, where displays are less than 5 in diagonal length and bandwidth availability is low. The proposed method exploits the mismatch between the video resolution, display resolution, and visual acuity of users. This method has very low impact on subjective quality while reducing encoding time for low bitrate environments. The proposed method reduces both encoding time and bitrate; the results show an average of 21.7% decrease in encoding time, 13.4% decrease in average bitrate, and very low impact on subjective quality ratings. The proposed methods can be adapted to decrease encoding time and bitrate in devices with lower resolution displays.

High Efficiency Video Coding (HEVC) is the new video coding standard recently approved by ISO and ITU. HEVC allows a bit rate reduction greater than a 50% with respect to its predecessor, the H.264/AVC, offering the same perceptual quality, by means of a set of new tools that have been introduced. Compared with the current state of the art in image coding, such as JPEG, JPEG2000 or the new JPEG XR, the new Intra Frame coding performs a high compression process in the "All-Intra" mode. All these improvements are at expense of a high computational cost, making it considerably difficult to implement in real time. Hence, this paper presents a mechanism that can be used by the RDO algorithm to select the optimal coding block size for Intra-Prediction, by using a data mining classifier, based on a previous training. Experimental results show that the proposed algorithm can achieve a 30% of Time Savings over a wide range of high resolution sequences (Class A, B and F), with a negligible loss of coding efficiency.

This is a MM'14 Workshop Summary Abstract for PIVP'14 - 1st International Workshop on Perception Inspired Video Processing. Workshop provided a venue for researchers involved with perceptual video processing and coding to present their current work and to discuss future directions. The workshop program consisted of a keynote talk, oral presentations of full papers and interactive poster session for the short papers. All participants took part in the discussion panel at the end of workshop, exchanging ideas and suggestions for future work.

We present a method that employs human visual system characteristics for perceptually lossless optimization of video coding. Algorithm is robust and applicable for any modern hybrid coder. We decided to evaluate it using the most recent standard (HEVC/H.265) in order to show potential benefits for future applications. Experiments show savings of up to 8% in bitrate without any loss in quality as confirmed by two sets of subjective experiments using DSCQS and ACJ methodology.

This paper presents a system and methodology for crowdsourcing subjective quality evaluations. The goal of this work is to study the feasibility of fast, inexpensive, and reliable subjective quality evaluations with crowdsourcing. We present a web-based system for conducting subjective quality evaluations using crowdsourcing platform. The results were evaluated by comparison with the formal subjective tests conducted by MPEG as a part of video standardization process. High quality compressed videos with known MOS values are used as references in subjective evaluations instead of the original lossless videos to address the bandwidth limitations. The results show that crowdsourcing can be a reliable platform for subjective quality evaluations.

This paper summarizes our work on optimization of video content compression in the context of the end user. I will briefly explain motivation and novelty of the work that presents the bulk of my dissertation research. Preliminary experiments with promising results are encouraging further steps that should lead to completion of a model for video compression that uses most of available bits for the video content that is actually seen. By this we mean the content that user attends to. Everything else is coded with much less bits, leading to significant savings compared to state-of-the-art coding techniques. Our work should be regarded as extension and not replacement of the hybrid coding paradigm.