Sung In Cho

Scene Change Detection Using Multiple Histograms for Motion-Compensated Frame Rate Up-Conversion

In this paper, we propose a new scene change detection method using multiple histograms. The proposed method produces multiple histograms of split blocks for consecutive frames. Then, it computes the optimal threshold value using automatic thresholding based on the Otsu method and decides whether the scene change occurs or not using the difference between threshold values in successive frames. In the experiments for the subjective evaluation, the proposed method correctly identified the scene change, thereby preventing interpolated frames with poor image quality due to block artifacts. In the objective evaluation using the F1 score, the proposed method improved the accuracy of the scene change detection by up to 0.461 compared with the benchmark methods.

Image Quality-Aware Backlight Dimming With Color and Detail Enhancement Techniques

In this paper, we propose an advanced backlight dimming technique that preserves the quality of color and details in images even when the backlight luminance of liquid crystal display (LCD) devices is lowered. The proposed backlight dimming technique consists of the following two steps: backlight luminance level selection and pixel compensation. In the first step, to reduce power consumption, the proposed approach selects an optimal level of backlight luminance for a given image based on image quality evaluation that considers the peak signal-to-noise ratio (PSNR) and color distortion level. In the second step, it adaptively adjusts the RGB ratio depending on image details, thereby enhancing image color and details, which are degraded by the lowered backlight luminance level calculated in the first step. The simulation results showed that the proposed method successfully selected the optimal backlight luminance level and prevented severe color distortion, while the benchmark method induced severe color distortion in some images. In addition, for the same backlight luminance level, pixel compensation in the proposed method reduced color difference for color distortion evaluation and the loss rate of edge strength, which showed detail loss by up to 3.58% and 40.55%, compared to benchmark methods, respectively.

Two-step local dimming for image quality preservation in LCD displays

This paper presents a new approach to local dimming for liquid crystal display (LCD) devices. Local dimming is a very important technique that LCD uses to reduce its power consumption, and is being increasingly adapted in wide display applications. The proposed method determines the backlight luminance level of each block by performing the following two steps. In the first step, it dynamically selects the initial backlight luminance level of each block considering the peak signal-to-noise ratio (PSNR). In the second step, the backlight luminance level of each block is refined by considering the light coming from neighboring blocks and the relations between the current block and its neighboring blocks. In the experiments, the proposed method improves the image quality of the processed images by up to 8 dB in terms of PSNR while consuming less power, when compared to benchmark methods.

Dictionary-based anisotropic diffusion for noise reduction

This paper presents an anisotropic diffusion-based approach to noise reduction, which utilizes a pre-trained dictionary for diffusivity determination. The proposed method involves off-line and on-line processing steps. For off-line processing, a multiscale region analysis that effectively separates the structure information from image noise is proposed. Using multiscale region analysis, the proposed approach classifies local regions and constructs a dictionary of several patch classes. Further, this paper presents a dictionary-based diffusivity determination that exhibits enhanced performance of anisotropic diffusion. In addition, we propose a single-pass adaptive smoothing that uses a diffusion path-based kernel, which is derived from iterative anisotropic diffusion operations. By using single-pass adaptive smoothing for both off-line and on-line processing, the proposed method is able to avoid the use of expensive iterative region analysis. In on-line processing, the proposed approach classifies input image patches using multiscale region analysis. It subsequently selects the diffusion threshold with the highest matching ratio from the dictionary for each region. Finally, single-pass adaptive smoothing is performed with the selected diffusion threshold. Simulations show that the proposed method outperforms benchmark methods by significantly enhancing the peak signal-to-noise ratio and structural similarity indexes.

Human perception-based image segmentation using optimising of colour quantisation

This study presents an advanced histogram-based image segmentation method that enhances image segmentation quality, while greatly reducing the computational complexity. Unlike existing histogram-based methods, the authors optimise the size of bins in the colour histogram by using human perception-based colour quantisation and the clustering centroids are selected effectively without using a complex process. Additionally, an over-segmentation removal technique based on connected-component labelling is employed. This improves the segmentation quality by connectivity analysis. A comparison between the experimental results on the Berkeley Segmentation Dataset by the proposed method and the benchmark methods demonstrated that the proposed method enhanced the segmentation quality by improving the Probabilistic Rand Index and the Segmentation Covering values compared with those of the benchmark methods. The computation time using the proposed method is reduced by up to 91.63% compared with the computation time using benchmark methods.

Image segmentation using linked mean-shift vectors and its implementation on GPU

This paper proposes a new approach to meanshift- based image segmentation that uses a non-iterative process to determine the maxima of the underlying density, which are called modes. To identify the mode, the proposed approach performs a mean-shift process on each pixel only once, and uses the resulting mean-shift vectors to construct links for the pairs of pixels, instead of iteratively performing the mean-shift process. Then, it groups the pixels of the same mode, connected through the links, into the same cluster. Although the proposed approach performs the mean-shift process only once, it provides comparable segmentation quality to the conventional approaches. In experiments using benchmark images, the processing time was reduced to a quarter, while probabilistic rand index and segmentation covering were well maintained; they were degraded by only 0.38% and 1.87%, respectively. Furthermore, the proposed algorithm improves the locality of the required data and compute-intensity of the algorithm, which are important factors for utilizing the GPU effectively. The proposed algorithm, when implemented on a GPU, improved the processing speed by over 75 times compared to implementation on a CPU, while the conventional approach was accelerated by about 15 times.

Multi-histogram based scene change detection for frame rate up-conversion

In this paper, we propose a new scene change detection method based on multi-histogram for frame rate up-conversion. The proposed method manages multiple per-block histograms to extract the locality of image change. Thus, it can detect local scene change as well as global scene change between frames. Experiments show that the proposed method improves by 14.05dB the image quality of the interpolated frame with the local scene change.

Image segmentation using linked mean-shift vectors for SIMD architecture

This paper presents a new mean-shift based segmentation algorithm for single instruction multiple data (SIMD) architecture. A standard mean-shift algorithm has different number of computations for each pixel because the endpoints of each pixels iteration process are different, thus a standard mean-shift algorithm is hard to be accelerated using SIMD architecture. The proposed algorithm, however, equalizes the number of computations for each pixel by constructing links between pixels using their first mean-shift vectors without iteration process. It makes the proposed algorithm more suitable for the SIMD architecture without a complicated scheduling module. Experimental results using the Berkeley segmentation dataset show the proposed algorithm successfully equalizes the number of computations with reasonable segmentation quality.

Accuracy enhancement of image segmentation using adaptive anisotropic diffusion

This paper proposes a new pre-processing method to enhance accuracy of image segmentation. The proposed method produces a de-textured image which gives appropriate help to improve the segmentation quality when the existing segmentation method, histogram-based clustering, is applied on the simplified image. For obtaining this simplified image, we perform the de-texturing using an adaptive anisotropic diffusion model. Then, the histogram-based clustering is performed on the de-textured image to obtain segmentation results. In the experiments the Berkeley Segmentation Dataset, probabilistic rand index (PRI) and segmentation covering (SC) values are used for evaluating the segmentation quality. Experimental results showed that the segmentation accuracy of the histogram-based clustering was improved by using pre-processing in terms of average PRI and SC values by up to 0.86%, 14%, respectively.

Foreground-based depth map generation for 2D-to-3D conversion

This paper proposes a foreground-based approach to generating a depth map which will be used for 2D-to-3D conversion. For a given input image, the proposed approach determines if the image is an object-view (OV) scene or a non-object-view (NOV) scene, depending on the existence of foreground objects which are clearly distinguishable from the background. If the input image is an OV scene, the proposed approach extracts a foreground using block-wise background modeling and performs segmentation using adaptive background region selection and color modeling. Then, it performs segment-wise depth merging and cross bilateral filtering (CBF) to generate a final depth map. On the other hand, for the NOV scene, the proposed approach uses a conventional color-based depth map generation method [9] which has simple operations but provides a 3D depth map of good quality. Human beings are usually more sensitive to depth map quality, and 3D images, for OV scenes than for NOV scenes. With the proposed approach, it is possible to improve the quality of a depth map for OV scenes than using the conventional methods only. The performance of the proposed approach was evaluated through the subjective evaluation after 2D-to-3D conversion using a 3D display, and the proposed one provided the best depth quality and visual comfort among the benchmark methods.