Search

Motion vectors histogram (MVH) is an effective feature vector in video indexing and analysis applications. One of the challenges in the use of motion vector histogram is that it is not robust to video aspect ratio (AR). As a result, it is not suitable for comparing videos with different ARs. In this paper a feature vector based on motion vectors in H.264/AVC is introduced which is robust to variations in aspect ratio. This feature vector, namely scaled motion vector histogram (SMVH), has superior performance compared to common MVH and experimental results indicates that SMVH achieves on average 10% improvement in similarity detection of videos with different ARs compared with MVH.

Intra coding in the High Efficiency Video Coding (HEVC) standard can significantly improve the compression efficiency but increasing the number of intra prediction modes and also higher number of Coding Unit (CU) sizes in the HEVC standard imposes much higher computational load compared with intra prediction in H.264/AVC, which substantially is computational intensive. To reduce the intra coding complexity in HEVC, this paper presents an early termination algorithm for intra prediction. The proposed method is based on the fact that a homogenous region can be predicted with larger CUs. As a result, when the proposed smoothness parameter is lower than a predefined threshold, only the prediction modes in the current CU are evaluated. Experimental results indicate that the proposed algorithm can provide on average 27.5% time saving with only 0.4% BD-rate loss whereas it maintains the same coding video quality compared with HEVC reference software, HM15.0, in all intra-main configuration.

In this paper, a texture based retrieval method is proposed in HEVC compressed domain. The proposed method can be used for either video retrieval of HEVC coded videos or image retrieval of HEVC I-frame coded images. In I-frame coding various prediction modes are utilized to spatially predict pixels of a block. The selected prediction mode and the size of prediction unit for coding of a block indicate the way, in which the pixels of the block are related to their neighboring pixels and they can considered as texture features. Thus, we measured the texture similarity of coded I-frames in HEVC based on the similarity of the prediction modes and block size histograms. This retrieval method achieves 0.34 ANMRR in videos with high resolution, which is better compared to the other compressed domain retrieval method based on previous standard H.264/AVC that have achieved 0.45 ANMRR in the same image database.

In order to achieve higher compression performance the amendment called Fidelity Range Extension (FRExt) was added to the H.264 Advanced Video Coding (AVC) standard which uses both 4×4 and 8×8 integer discrete cosine transform (DCT) adaptively in the high profiles. This led to additional complexity of the initial version of the H.264/AVC encoder which had substantially high computational complexity. In this paper we propose a new algorithm which reduces the computational complexity for software implementation of horizontal 8×8 Integer DCT by more than 25 percent. Simulation results indicate 22 percent reduction in the computation time by using the proposed algorithm.

In this paper, a new method for extracting Farsi text from a video in the wavelet domain is proposed. In an image including text, due to the complexity of the background and the quick changes of contrast in the text areas, the energy intensity variation is high. Therefore, by exploiting this property a method is proposed for extracting text. Moreover, for increasing the accuracy of the proposed method, some motion characteristics of the movies have been used. The proposed method is able to localize both Farsi and Latin Text in video with a high performance. The experimental results confirms the effectiveness of the proposed method.

One of the most important aspects of image quality is contrast and this paper presents a new hybrid method for contrast enhancement. The proposed method is a combination of two basic contrast enhancement methods i.e. transform and histogram. At first we apply Nonsubsampled Contourlet Transform (NSCT) on the source image, then NSCT coefficients are mapped to fuzzy domain and modified by a mapping function in fuzzy domain. After transforming the modified membership values from fuzzy domain into frequency domain, the enhanced image is reconstructed from the modified NSCT coefficients by inverse NSCT. Finally, histogram of the image is equalized by Contrast Limited Adaptive Histogram Equalization (CLAHE). The experimental results show that our method can achieve higher improvement than both Contourlet Transform (CT) and Histogram Equalization (HE) contrast enhancement methods.

Binary Arithmetic Coding (BAC) is among the techniques used in H.264 video coding standard to improve the coding efficiency. BAC includes an iterative process of renormalization with up to seven iterations for coding each symbol. Since BAC is also a computational intensive unit in H.264 encoder, various hardware realizations have been proposed for it in the literature. In this paper, we propose a hardware implementation for BAC, which uses lookup table to avoid the iterative coding process and achieves coding rate of one symbol per clock at 260 MHz clock rate. Post synthesize simulation results indicate that the proposed architecture is a resource and speed efficient hardware for H.264 binary arithmetic encoder.

Content based image retrieval is a challenging issue in management of existing large digital image libraries and databases. The accuracy of image retrieval methods is subject to effective extraction of image features such as color, texture, and shape. In this paper we propose a new image retrieval method using contourlet transform coefficients. We use the properties of contourlet coefficients to assign the normal distribution function to the distribution of coefficients in each sub-band. The assigned normal distribution functions are used to extract the texture feature vector at the next stage. Simulation results indicate that the proposed method outperforms other conventional texture image retrieval methods such as, Gabor filter and wavelet transform. Moreover, this method shows a noticeable higher performance compared to another contourlet based method.

In this paper we propose a new indexing method for comparing image contents in the JPEG2000 compressed domain. Comparing image contents of the JPEG2000 coded images in the pixel domain requires image decompression, which imposes intensive computational processes of the inverse discrete wavelet transform and the arithmetic decoding. On the other hand, the first decoding stage of the JPEG2000 standard is packet header decoding, which is a simple process, but provides valuable information about the code blocks in the packet. In this paper we use packet header information for JPEG2000 image indexing. The proposed method exploits full packet header information including the number of non-zero bit-planes, the number of coding passes and the code block length for indexing the JPEG2000 compressed images. Experimental results show that the proposed method provides a better performance compared to other JPEG2000 compressed domain indexing techniques and even outperforms the pixel-based image indexing techniques such as the Gabor filter.

In this paper we introduce a compressed domain texture retrieval method for I-frame coded pictures in the H.264 standard and study its robustness against the image and coding parameter changes. The proposed image retrieval method uses the intra mode prediction histogram of the I-frame coded pictures as the feature vector. The simulation results show that this is a very efficient compressed domain image retrieval method and is also robust to various image and coding parameter changes. The robustness of the proposed feature vector indicates that the proposed method can be used in a wide range of compressed domain image and video retrieval applications.