Tien-Ying Kuo

Fast macroblock partition prediction for H.264/AVC

Variable block size motion compensation has been adopted by the emerging video coding standard H.264. For a macroblock, motion estimation on each block type has to be tested in H.264 to find the best block size in use. It results in high computational complexity when examining all possible block types. Thus, a fast block type prediction is proposed to reduce the complexity. The proposed method predicts a suitable macroblock partition according to the motion field distribution and correlation within the macroblock. Therefore, only the predicted mode should be tested in motion estimation

A novel image binarization method using hybrid thresholding

In this paper, we proposed a simple yet effective image binarization method to work with various kinds of images, such as images with the color bleeding characters, or with non-uniform background exposed to poor ambient light, and so forth. Our approach is based on a hybrid color quantization process, which adaptively takes the global and local image characteristics into account; therefore, it can deal with complex images. The experiment result shows that our proposed hybrid method outperforms the literature methods.

Fast mode decision for non-anchor picture in multiview video coding

In this work, we aim to develop a fast block mode decision in the multiview video coding (MVC) for the non-anchor pictures, especially with motion skip (MS) mode enabled to further exploit the inter-view correlations for higher coding efficiency. To combat with the complexity increase with the additional adopted MS mode, one primary contribution of this work is to speed up the MS mode decision by smartly choosing or skipping it. Experimental results show that our proposed method can gain the MVC coding speed while keeping similar rate-distortion performances on the non-anchor pictures.

Hybrid temporal-spatial error concealment technique for video communications

This work proposes a hybrid temporal and spatial error concealment technique. We first present, in the temporal domain, an improved boundary match algorithm to recover lost motion vectors, and it preserves better edge continuity on the boundary of the recovered block. In our spatial error concealment, we reduce the high computational complexity of the method of the weighted Markov random field (MRF) to make it feasible in real applications. Finally we hybridize our temporal and spatial error concealments together. The hybrid criterion is based on the motion fields of the damaged frame and the similarities of two contiguous frames. Experimental results show that our proposed hybrid error concealment algorithm can provide good recovery of missing blocks.

Variable frame skipping scheme based on estimated quality of non-coded frames at decoder for real-time block-based video coding

This paper proposes an encoder employing variable frame skipping (VFS) to improve the video quality in low bit rate channels. The basic idea of the VFS mechanism is to decide and skip a suitable, non-fixed number of frames in the temporal domain to reduce bit usage. The saved bits can be allocated to enhance the spatial quality of the video. In the literature, several methods for frameskip decision have been proposed, but most of them only consider the similarities between neighboring coded frames as the decision criteria. Our proposed method takes into account the reconstruction of the skipped frames using motion-compensated frame interpolation at the decoder. The proposed VFS models the reconstructed quality of the skipped frame and, therefore, can provide a fast estimate to the frameskip at the encoder. The proposed VFS can decide the frameskip in real time, and its encoded video has better spatial-temporal bit allocation

Fast sample adaptive offset encoding for HEVC

This paper presents a rapid method for predicting the best type of Sample Adaptive Offset (SAO) in the HEVC video encoder. The main idea of this work is to utilize the available motion vector and frame characteristics, and to reuse the information given by the reference picture and the neighboring blocks to accelerate the SAO decision while keeping good coding efficiency.

Image forgery detection for region duplication tampering

This paper proposes an automatic tampering detection algorithm to detect the region duplication forgery on a digital image. This algorithm uses dual spare features and structure similarity to enhance the detection performance of the region duplication. We also create a local outlier detector to recognize the intrinsic repeated elements to avoid the false positive detection. For a quantitative performance analysis, the forgery images are automatically generated the region duplication with geometric and brightness distortion by the tampering generator. The experimental results show our proposed algorithm is more robust and effective than similar works in literature.

Depth estimation from a monocular outdoor image

We present an algorithm to estimate 3D depth from a single monocular outdoor image based on the multiresolution analysis of the various cues including edge, hue, saturation, vertical depth prior, and dark channel prior. Our experimental results show a huge amount of computational savings over literature work while keeping our depth map close to the ground truth.

Multi-Shot Framework with Preloading Architecture for Low-Latency MHP Application Delivery

This paper proposes a Multimedia Home Platform (MHP) multi-shot application framework to reduce the delivery latency from the headend to the MHP terminal. The framework is conformed to the MHP standard and is optimized based on the characteristics of DSM-CC object carousel in data broadcasting. This architecture makes a division of application resources into shot units, optimizes the application start-up time by repeating the very initial shot in the carousel, and introduces a pre-loading method to cache the next shot candidates in advance. Therefore, this framework can allocate resources in the receiver more efficiently and significantly improve the viewers responsiveness.

An improved exemplar-based image repairing algorithm

Image inpainting is a technique to repair the damaged part of images or to remove unwanted artificial post-production effects on them. The image after inpainting should be restored to a state which looks like it has never been altered before. In this paper, we proposed an image inpainting technique with a simple mechanism to divide patches of an image into different priority levels. While most literature works need to inpaint these levels in a restricted order, we implemented a scheduling mechanism which frees up the inpainting process from this rigid repair order, allowing for better performances.