Kang-Sun Choi

New autofocusing technique using the frequency selective weighted median filter for video cameras

Most conventional autofocusing techniques based on the gradient estimator are very sensitive to noise. A new autofocusing technique which is resistive to noise generated by the CCD of video cameras is proposed. In the proposed scheme, the frequency selective weighted median (FSWM) filter is utilized to estimate the degree of focus and the fast hill-climbing search (HCS) strategy is exploited to determine the best focused image. Since the FSWM filter can not only extract high frequency components from the image, but also eliminate impulsive noise, the proposed autofocusing method employing the FSWM criterion function can estimate the degree of focus precisely. Furthermore, the proposed real-time HCS algorithm enables the video camera to continuously focus on dynamic images. Experimental results demonstrate that the proposed technique outperforms existing techniques by enhancing the accuracy of the focus value of the video camera without the influence of noise.

Real-Time System for Counting the Number of Passing People Using a Single Camera

In this paper, we propose a real-time people counting system with a single camera for security inside the building. The camera is hung from the ceiling of the door so that the image data of the passing people are not fully overlapped. The implemented system recognizes people movement along various directions. To track people even when their images are partially overlapped, the proposed system analyzes the captured image and estimates important feature information. Then each bounding box enclosing each person in the tracking region are tracked. The convex hull of each individual obtained by the proposed algorithm provides more accurate tracking information. Through extensive tests in various environments, the proposed system exhibits the correct people counting results of over 96%.

Wavelet based seam carving for content-aware image resizing

In this paper, a novel content-aware image resizing method based on wavelet analysis is proposed. We estimate the local energy map of an image by weighing its multiscale subbands appropriately. Based on the energy map, the image is resized by repeatedly carving out or inserting in a connected path of pixels which is least significant in terms of the energy. Since wavelet analysis is similar to the way the human visual system operates, the obtained energy map reflects human perception with fidelity, and thus, the semantic information in the image can be preserved faithfully in the resizing process. The experimental results show that the proposed method produces higher subjective quality images than scaling and conventional content-aware image resizing techniques.

Real-Time Inspection System for Printed Circuit Boards

The PCB inspection system presented in this paper can detect defects including the breaks in the wires and short circuit and is significantly faster when compared to the existing techniques. The proposed inspection method is based on referential matching between the stored reference image and the test (observed) image. Block matching is performed to solve the misalignment in referential matching. In order to reduce the computational complexity, we implement the proposed algorithm using the single instruction multiple data (SIMD) instructions, so called SSE2.

Improved seam carving using a modified energy function based on wavelet decomposition

In this paper, a novel content-aware image resizing technique, perceptually improved seam carving with modified energy function, is proposed. The conventional seam carving method using a gradient-based energy function may cause distortion of deformed object shape depending on the layout and amount of contents. However, the proposed method preserves content shape with high fidelity by exploiting a wavelet-based energy function which represents image features effectively within the scale-space framework. To prevent excessively distorting the images, a switching scheme between seam carving and resampling is proposed based on the energy distribution. Experimental results show that the proposed method can achieve higher subjective image quality than the existing methods including scaling and conventional seam carving.

Real-time DSP implementation of motion-JPEG2000 using overlapped block transferring and parallel-pass methods

This paper presents a real-time implementation of Motion-JPEG2000 encoder using a fixed-point DSP chip. Among several modules in JPEG2000 encoder, the lifting algorithm for discrete wavelet transform (DWT) and the embedded block coding with optimized truncation (EBCOT) comprise more than 85% of the encoding complexity. Thus, it is very important to design and optimize these two modules in order to increase the performance of the hardware implementation. First we propose an overlapped block transferring (OBT) method that can significantly improve the performance of the lifting algorithm for DWT by increasing the cache hit rate. We show that the execution time of the lifting scheme can be further reduced by programming the DSP software using the single instruction multiple data (SIMD) instructions and the super scalar pipeline structure. Moreover, we introduce a parallel-pass method for fast implementation of EBCOT. This method reduces the processing time of EBCOT by processing the three coding passes of the same bit-plane in parallel. Experimental results show that our developed Motion-JPEG2000 DSP system meets the common requirement of the real-time video coding [30 frames/s (fps)] and is proven to be a practical and efficient DSP solution.

Fast content-aware image resizing scheme in the compressed domain

Since consumer devices have different resolution sizes of display panels, digital images are frequently required to be resized for each display adaptation. In this paper, we define DCT-based distortion measures to quantitatively evaluate both scaling distortion and shape distortion generated by downsizing. By utilizing the distortion measures, we propose a novel content-aware image retargeting algorithm. The proposed algorithm can not only satisfy demands of users without the loss of semantic information in images but also is computationally efficient since the digital images are typically stored in DCT-based compression format. Experimental results show that the proposed method can achieve higher subjective quality with significantly reduced computational loads as compared with existing methods.

Backlight power reduction using efficient image compensation for mobile devices

Since backlight dimming widely used for power reduction of mobile devices degrades the image quality in brightness and contrast, image enhancement is exploited to compensate the degradation. In this paper, we propose a novel backlight power reduction scheme using an efficient image compensation algorithm that scales discrete cosine transform (DCT) coefficients of image in the compressed domain. To reduce the computational burden of the conventional backlight dimming method where decompression is performed prior to image compensation in the spatial domain, we introduce a computationally efficient image compensation algorithm that processes only dominant DCT coefficients of images using a mapping function in the DCT domain. Experimental results indicate that our proposed method provides higher visual quality with reduced complexity as compared to conventional image compensation schemes.

Enhanced frame rate up-conversion method for UHD video

In this paper, a segment-based motion estimation (ME) method for frame rate up-conversion is proposed that can achieve better visual quality for next generation video formats. The proposed ME method can reduce the computational complexity of the ME drastically by adjusting the search range (SR) for each segment and adopting a new efficient distortion criterion, called the subsampled sum of absolute differences (SSAD). The proposed method does not degrade the motion accuracy by using the temporal consistency of the segment with large size. Experimental results show that the proposed algorithm outperforms conventional methods both objectively and subjectively with the significantly reduced computational complexity.

Low complexity content-aware video retargeting for mobile devices

In this paper, a low complexity content-aware video retargeting method suitable for mobile multimedia devices is presented. We propose two types of energy functions for intra- and inter-frames to obtain the pixel significance in the video frames. Since the energy functions are defined using only the compressed-domain information such as discrete cosine transform (DCT) coefficients and motion vectors (MVs), instead of fully decoded pixel domain information, the spatiotemporal energy map can be constructed without excessive computations. Based on the proposed energy function, we also introduce a fast seam carving method for mobile devices that carves out the one-dimensional (1-D) seams without annoying artifacts such as jitter and flicker, taking into account the object motion based on MVs. Experimental results show that the proposed video retargeting scheme can achieve high subjective quality with significantly reduced computational loads as compared with existing methods.