A major problem for the block-based discrete cosine transform (BDCT) techniques is that the decoded images, especially at very low bit rates, exhibit highly noticeable blocking artifacts near the block boundaries. In this paper, a new deblocking algorithm based on wavelet transform and Markov random field (MRF) is proposed. The blocking-artifacts image is firstly processed with a simple and effective wavelet-based deblocking algorithm using the two-scale wavelet scheme. Adaptive operators for different subbands can be computed to suppress blocking effects below the visual scope. It can make the image smoother, which provides advantage for the following MRF-based deblocking method. The proper threshold value for Huber function plays an important role in MRF method. The linear regression is used to adaptively estimate the threshold value. Experimental results show that the new algorithm has low computational cost and achieves improved image quality in both subjective and objective measurement
Based on the searching analysis of computer, the lifting coefficients of the 97 biorthogonal wavelet filter pair of CDF are expressed in arithmetic operation of tow shift registers. A arithmetic operation of shift registers are used to replace the multiplier with complicated architecture operating on floating-point. So it is more simpler and faster, and can be implemented by ASIC easily. In the end, the analysis of some simpler algorithms and the configuration of arithmetic shift are given.
A method of digital signal processing based on FPGA chip is presented. The system function is described by VHDL and circuit diagram, compiled and synthesized with MAX+plus II, and implemented in a FPGA chip in series of FLEX10K from Altera company. In this way base-6 FFT algorithm is implemented, and the simulation results are also presented. The Fast Fourier Transform(FFT) algorithm running on FPGA chip has the advantages of high speed and excellent anti-disturb capability, and the FFT algorithm designed by VHDL based on IP core can be used repeatedly so as to improve the design efficiency.
When a high compression ratio is required, the block discrete cosine transform (BDCT) decoded image suffers from the blocking effect: boundaries between adjacent image blocks visible for low bit rates. A blocking effect reduction algorithm based on DCT transform and Markov random field (MRF) is proposed to reduce blocking artifacts as many as possible, while preserving edge information. In this paper, the characteristic of human visual system (HVS) is sufficiently utilized. A visibility function of blocking artifacts is proposed based on the activity masking and brightness masking properties of HVS. The blocks which visibility of the blocking artifacts is less than the threshold do not need any processing. The proper threshold value for Huber function plays an important role in MRF method. The linear regression is used to adaptively estimate the threshold value. Experimental results show the proposed algorithm can significantly reduce blocking artifacts while preserving edge and texture information
In order to achieve the real-time property for video coding, a fast and adaptive algorithm based on starting search point prediction and early-termination strategy is proposed. It analyzes center-bias property and spatial correlation property of motion vector field, and utilizes the respective characteristics of block based gradient descent search (BBGDS) and adaptive rood pattern search (ARPS) algorithm. The proposed algorithm adaptively chooses different searching strategies according to the type of the image, makes full use of the cross image motion vector distribution characteristics and optimizes the traditional ARPS algorithm. The experimental results show that the proposed algorithm is about 2.3∼9.2 times faster than Diamond Search (DS), 1.2∼4.0 times than ARPS. The algorithm can meet the real-time demand without reducing the image quality.
A novel fast motion estimation algorithm called M-DAHS based on motion vector field,adaptive direction and half-pixel search was proposed.According to center-bias property and spatial-temporal correlation of the vector field,the threshold was applied to stop the stationary block from searching.The initial search point and strategy were adaptively selected based on motion type for non-stationary block.The search template has strong direction adaptability.The line-diamond search was used for the blocks with low motion activity,and the hexagon-diamond search for that with middle or high motion.After searching all the pixels,a cross priority search pattern was utilized for half-pixel search.Experimental results show that the proposed algorithm has good performance with high search speed and high precision,while the distortion is almost the same as full search algorithm.
Real-time semantic segmentation is widely used in autonomous driving and robotics. Most previous networks achieved great accuracy based on a complicated model involving mass computing. The existing lightweight networks generally reduce the parameter sizes by sacrificing the segmentation accuracy. It is critical to balance the parameters and accuracy for real-time semantic segmentation. In this article, we propose a lightweight multiscale-feature-fusion network (LMFFNet) mainly composed of three types of components: split-extract-merge bottleneck (SEM-B) block, feature fusion module (FFM), and multiscale attention decoder (MAD), where the SEM-B block extracts sufficient features with fewer parameters. FFMs fuse multiscale semantic features to effectively improve the segmentation accuracy and the MAD well recovers the details of the input images through the attention mechanism. Without pretraining, LMFFNet-3-8 achieves 75.1% mean intersection over union (mIoU) with 1.4 M parameters at 118.9 frames/s using RTX 3090 GPU. More experiments are investigated extensively on various resolutions on other three datasets of CamVid, KITTI, and WildDash2. The experiments verify that the proposed LMFFNet model makes a decent tradeoff between segmentation accuracy and inference speed for real-time tasks. The source code is publicly available at https://github.com/Greak-1124/LMFFNet.
In this paper, a new method for motion target detection by background subtraction and update is proposed in order to modify the deficiencies of the traditional detection methods. In this scheme, the dynamic threshold can make up for the shortcomings of misjudge by fixed threshold background subtraction so as to extract most motion pixels. The local background updating strategy based on double threshold counting can promptly update background and overcome the shortcomings of the traditional background updating methods. The morphological processing method is applied to remove various noises to polish the image edge. The experimental results show that the proposed method has the low computational complexity, and can well detect the motion target under the very complex background.
To improve the performance of LMS adaptive algorithm, in this paper a new variable step-size LMS algorithm is proposed based on the analysis of some variable step-size algorithms. Through establishing a new nonlinear relationship between the step size and the error, the algorithm eliminates the irrelevant noise and improves the convergence rate to obtain a better stability. And the computer simulation results are consistent with the theoretical analysis, which confirmed that the algorithm is superior to other algorithms on convergence rate, tracking speed and steady-state error.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)