LCD Motion Blur Analysis

Introduction

Liquid crystal display (LCD) is the most popular display device in the market due to its low cost, low power consumption, high resolution and high contrast. However, LCDs are also known for its slow response time, which in turn causes motion blurs. In this project, we study several fundamental properties of LCDs. Our goal is to provide both theoretical and experimental justifications to some methods employed in the industry.

LCD motion blur reduction is an inverse problem: Given a target signal (sharp), we need to pre-distort it so that the pre-distorted signal can compensate the liquid crystal system, which is a low pass system. The schematic diagram shown in Figure 1 illustrates the two important building blocks in this study. First, the impulse response of the system must be thoroughly studied, for otherwise it is impossible design methodologies to overcome the LCD motion blur. Second, provided a model of the LCD system, algorithms should be developed to reduce motion blur.

Figure 1 LCD motion blur reduction problem. Our goal is to understand the system response and determine the unknown signal.

Publication

Stanley H. Chan and Truong Q. Nguyen, "LCD Motion Blur: Modeling, Analysis and Algorithm," IEEE Transactions on Image Processing, vol.20, issue 8, pp.2352-2365, Aug 2011.
- Supplementary videos ( link )
Stanley H. Chan, Thomas X. Wu and and Truong Q. Nguyen, "Comparisons of two frame rate conversion schemes for reducing LCD motion blurs," IEEE Signal Processing Letter, vol. 17, issue 9, pp.783-786, 2010.
- Supplementary videos ( link )
Stanley H. Chan and Truong Q. Nguyen, "LCD Motion Blur Modeling and Simulation," in Proceedings of IEEE International Conference on Multimedia and Exposition, pp.400-405, Singapore, 2010. (Oral presentation)
- Supplementary videos ( link )
Stanley H. Chan and Truong Q. Nguyen, "Fast LCD Motion Deblurring by Decimation and Optimization," in Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pp.1201-1204, April 2009 Taipei. (Poster presentation)
- Supplementary videos ( link )

A. Liquid Crystal Step Response

In this project, we study the step response of a liquid crystal (LC). From step response, we can then derive the impulse response of a liquid crystal and hence the transfer function of the LCD. Our derivation is based on the solution of Erickson-Leslie equation. Compared to the conventional resistor-capacitor (RC) approximation, and uniform averaging approximation, our model is able to capture several important characteristics of LC's step response. These includes fast rise time and slow fall time, existence of pre-tilt angle, and dependence to gray level transition. [Chan Wu 2010]

Project Page: Here

Figure 2 Step responses of a liquid crystal using three different models. The Erickson-Leslie model (blue colored curve) is able to capture the characteristics of a liquid crystal that the other two models are not able to.

B. Comparisons of two 240Hz LCDs

In this project, we compare two popular frame rate up conversion (FRUC) schemes for 240Hz LCD devices. The first method is a direct 4x conversion from 60Hz to 240Hz by motion estimation and motion compensation (full ME/MC). The second method is a 2x ME/MC from 60Hz to 120Hz, followed by black data insertion. From the physics point of view, we show that in time domain, full frame insertion method produces smoother motion and hence is better. However, in spatial domain, neither of the two methods is better because the spatial profile of black data insertion method is time dependent. [Chan Wu 2010]

Project Page: Here

Figure 3 Comparisons between two commonly used frame rate up conversion methods for 60Hz to 240Hz.

C. Spatio-Temporal Equivalence

Motion blur is a temporal subject, yet it is closely related to spatial averaging. In this project, we establish the equivalence between the temporal averaging and the spatial averaging for digital video sequences. We prove that if the relative motion between frames is not dramatic, then the temporal averaging can be accurately approximated by a spatial averaging. Our derivation is valid for both global and local motions. The equivalence allows us to construct spatial operators to simulate the temporal motion blur.

Proof: Here

Figure 4 Illustration of the spatial-temporal equivalence. To evaluate the temporal integration, we first fix a position (x0, y0) and consider the pixel values at different times t = 0, . . . , 3. The average is taken over the time, so it is the average across the four marked pixels on the right hand side. However, since these four frames are identical to each other (after motion compensation), we can evaluate the temporal average by averaging four adjacent pixels (in spatial domain).

D. Maximum Filter Length

Motion compensated inverse filter (MCIF) is a standard methodology of reducing LCD motion blur from a signal processing stand point of view. MCIF is a finite impulse response (FIR) filter design problem. This inverse filter design depends heavily on the underlying forward blur operator. If the forward blur operator is stronger, then inverse filter has to be more aggressive. However, more aggressive inverse filter causes more artifacts. In this project, we pose a constraints to the length of the forward blur operator. We show, through human subjects, that if the length of the forward blur operator is longer than certain threshold, it is better not to use an aggressive inverse filter. Rather, we should use a mild inverse filter. [Chan 2010] , [Chan 2011]

Video Demo: Here

Figure 5 The effects of increasing the length of the FIR motion blur filter and design an inverse filter. As the length of the motion blur filter increases, more artifacts are generated. However, the perceived sharpness also increases. Our goal is to determine the optimal length of the FIR filter so that there is a balance between the perceived sharpness and the artifacts.

E. LCD Motion Deblurring Algorithm

In this project, we develop an algorithm for LCD motion blur reduction. Our methodology is to consider the problem as an inverse signal synthesis problem. We formulate the problem as a least-squares minimization problem subjected to spatial and temporal constraints. Then we develop a customized optimization algorithm to solve the minimization problem. [Chan Nguyen 2011] [Chan Nguyen 2009]

Figures below show the comparisons between the following three papers:

[1] M. Klompenhouwer and L. Velthoven, "Motion blur reduction for liquid crystal displays: Motion compensated inverse filtering," in Proceedings of SPIE-IS and T Electronic Imaging. SPIE, 2004.
[2] Shay Har-Noy and Truong Q. Nguyen, "LCD motion blur reduction: A signal processing approach," IEEE Transaction on Image Processing , vol. 17, pp. 117-125, Feb 2008.
Stanley H. Chan and Truong Q. Nguyen, "LCD Motion Blur: Modeling, Analysis and Algorithm," accepted to IEEE Transactions on Image Processing, 2011.

Video Demo: Here

Figure 6 Comparisons between several LCD motion blur reduction methods.

F. GPL1 - An Algorithm for Video Deblurring

GPL1 stands for Gradient Projection for L1 minimization. It is a sub-gradient projection minimization algorithm that can be applied to wide range of image and video restoration problems. Specification application of GPL1 is for spatially variant video deblurring, using spatially variant blur operators.

The minimization that GPL1 solves is the following anisotropic total variation regularlized least-squares: [Chan, Nguyen 2011]

Here, the operator H is the convolution matrix, D is a set of forward difference operators oriented at different angles, M is the motion compensation operator and f_{k-1} is the solution of the previous frame.

Blurred
Recovered by GPL1

Blurred
Recovered by GPL1

Patents

Stanley H. Chan and Truong Q. Nguyen. GPL1 - An Algorithm for Video Deblurring. Disclosed to UC San Diego. 2010.
Stanley H. Chan and Truong Q. Nguyen. LCD Deblurring using Numerical Optimization with Spatial Temporal Regularizations. Disclosed to UC San Diego. 2008.