2006 Vol. 31, No. 9
To remove the color or luminance differences within a single image and between multi-images in block area,an auto-dodging processing approach and the corresponding integrated processing flow are proposed.Its principle and processing flow are implemented in the development of the software GeoDodging4.0 and good effects are achieved in practical project applications.
A wavelet visible watermarking algorithm is proposed based on noise masking,then the improvement is made based on texture feature of watermark image.Based on the(above) idea,an image content holding strategy and the algorithm are proposed.Experimental results have showed that the proposed algorithms can ensure watermark visibility and the loss in host image become less.
The spatial information of the image is applied to image retrieval based on generalized image co-occurrence matrix.Experimental results show that the retrieval efficiency is improved largely compared with the classic co-occurrence matrix method.
In order to evaluate the halftoning algorithms and the corresponding halftones,a criterion is necessary.A series of halftonig algorithm-based methods are proposed,which(adapt) to the special properties of haltftoning algorithms.An adaptive method is proposed to evaluate the halftoning algorithms and the corresponding halftones,which is based on the statistical features of the residual image between the original image and the corresponding halftone on the retinal of human eye.
The descreen methods for color or grayscale images captured by scanner are generally blur arithmetic.But for halftone image generated by RIP the blur arithmetic is useless because the image is bi-level bitmap one so that it is of only 0 and 1 two levels.The descreen method is based on a very different arithmetic,it finds out the center of halftone cell and screen ruling parameters first,and then divides the image into halftone cell.By computing dot area percentage in each halftone cell and converting its value to the grayscale,the halftone dots would be eliminated.The computing results show that the descreening results dependent upon the resolution,screen ruling angle and frequency of the halftone image.The descreen effect of this method is reasonable good and a full color image can be composed from 4 RIPed separation images.The method can be used to digital soft proof based on the RIPed separation images.
How to use the digital image technology to auto-recognize traffic signs in MDCS(mobile data capturing system) is discussed.With kinds of image process to extract their feature and match them,the veracity of recognizing traffic signs can be improved.Therefore,increasing efficiency of traffic sign surveying and updating the period of surveying can be achieved.
An algorithm is presented based on curvature variance to calibrate the lens distortion of digital camera.At first,straight lines are extracted in the image while the actual line is slightly curved due to lens distortion.Therefore lines extracted are dimidiated and each line section is extracted again to detect the curvature caused by lens distortion.Then the curvature is used to calibrate lens distortion.It is proven by experiments that this algorithm comes out with satisfying result.
An novel,digital screening method is developed to adopt different halftone algorithms by image segmentation.Experiment shows that it can represent tones and details better on low-resolution devices in this way, and a better binary result is expected.
A new approach is presented for vehicle license plate location and rectification based on RGB chroma space in the complicated-background and different illumination images.According to the character of RGB chroma space,a vehicle plate location model was built through a series of operations in RGB chroma space.Experimental results show that the vehicle license plate location and rectification algorithms based on RGB chroma space can improve the whole performance of vehicle license plate recognition system greatly.
On the basis of the hybrid algorithm constructed by cluster analysis and principal component analysis a newer multi-spectral space for color representation is developed.When cluster analysis applied using self-organized mapping the spectral distributions in each cluster are substituted by the mean distribution of those in the corresponding cluster.Further,principal component analysis is implemented on the compressed spectra set.Using the eigen spectral distributions as base to spanning the wide-band multi-spectral space for color representation.Results of experiments show that this newer multi-spectral space can represent and reconstruct the spectral distribution of the original color in a high accuracy.
The deforming of landslides usually expressed in horizontal and vertical direction separately.Based on the X,Y,Z coordinates from a real monitoring project.A method is proposed to build 3D model of landslides by constructing triangulated irregular network(TIN) and extruding contour lines.The tiny movement of monitoring points becomes visible by zooming movement vectors,and the whole landslide deforming was replayed using rubber-band animation techniques.
A neural network method for the color conversion from CIE colorimetric values to RGB control signals for color digital hardcopy devices is presented.Based on the color transfer mechanism of the making equations,the color conversion can be performed by introducing the colorimetric density variables.A BP neural network is utilized to achieve the three-dimensional transformation from the synthetic to analytical colorimetric densities.And then the analytical colorimetric densities are translated to RGB values by three one-dimensional computations.Experiments were carried out on several color devices.Compared with the conventional 3D_LUT color conversion,the network method had a close mean and much more lower maximal errors.
Through analysis of the generation mechanism of compression distortion,a visual masking effect based adaptive image compression algorithm that can reduce the "smoothing effect" in low bit rate coding is proposed.For the reason that images should be modeled as non-stationary information resources,blocking wavelet transform is used.The decomposition levels of a block are determined by texture complexity of that block.Experiments by comparing with baseline JPEG2000 show that visual quality in low bit rate coding is improved significantly,and more weak information is reserved.
A quantitative analysis method for color gamut of digital original accurately is proposed.Tthrough relevant experiments various formation of image environment to digital original influence of the color gamut are compared,some achievements with value are obtained.
This research object mainly developed the raw materials—sericite,which replaced the input silica materials due to its characteristic properties.There are five coatings formulations were prepared,input silica was replaced by domestic sericite from 0% to 20% proportion,then uses these coatings and base paper with coater to make a novel digital imaging output paper.Lastly,color data evaluations and analysis on paper from inkjet printer are finished with color target chart through spectrophotometer.The outcome of this research found out that overall color gamut decreased,the brightness of shade area increased from 5% to 10% sericite proportion.But it initiated to turn back suddenly in 15% sericite proportion,overall brightness continued to increase,color gamut domain broadened out simultaneously.In 20% sericite proportion,the color gamut domain broadened out clearly more than 15%,and overall brightness was the widest than any proportions.The brightness of shade area increased in proportion to the sericite,which replaced silica.Color gamut domain turned back from 10% to 15% sericite proportion.
An experiment will be done with 9 color centers(216 color samples) and 20 observers to find out the mean of acceptability threshold of each color difference,and visual scale result is used to improve color difference by optimum l∶c weighting.The CIE94,CMC and CIE2000 color difference equations were used,ΔE94,ΔE_(cmc),DE2000 were compared with ΔEab.
Along with the flexography printing obtained the vigorous development,the researches on the dot gain and deforming of the flexor plates are become more and more important.The main factors that affect dot gain of flexography are analyzed,such as print pressure,screen line of plate,hardness of plate.The affecting curve is studied through the experiments.The programs and ways to control dot gain are obtained,these will help to(achieve) high flexography printing quality.
A new flat plane model is proposed using interpolations,which converts L~*a~*b~* color space to cmy color space and analyzes the accuracy of the transformation.The flat plane model of ink jet printing is compared with that of offset,which gives insight and feasible methods in order to establish the flat plane models for all kinds of printing means.
A printer model and human visual model based method on tone-dependent error diffusion(TDED) is proposed.First the printer model and HVS(human visual system) used in halftoning algorithm are analyzed.And based on the Floyd-Steinberg error diffusion algorithm,the tone-dependent weights matrix is trained on the different tones using an error within a human visual system.In the experiment,the printer is used to test the result and compare the quality of the halftone image created by the new algorithm,and both them show very good halftoning quality.
Began with introducing the stomatic cubage of paper capillary and absorption to liquid of paper capillary,a method to study the distribution of paper's capillary is advanced,and based on that,the penetrating phenomena and penetrating depth are analyzed.
The fragile digital watermarking scheme is presented. The results of experiments reveal that the scheme presented here is greatly practical,with an excellent ability to locate juggling actions and of high security,as the same time,because of non-overlapping,time complexity is good than others algorithm.
An improved Red/Blue sweep line algorithm is preserted for connected line segment intersection in GIS.First the algorithm breaks down the connected segments into monotone chains.By filtration for monotone chains,those monotone chains which can not generate intersection points are removed.Then algorithm computes intersection points among rest monotone chains by traditional Red/Blue sweep line algorithm,which is based on isolated line segments.Both theoretic analysis and experiments show that our algorithm performs more efficiently than the traditional Red/Blue sweep line algorithm does.
The load balanced technology of VOD cluster server is analyzed,and a linear transformation genetic algorithm is put forward.It first generates a fitness function through weighted least connection algorithm,and than generates a new group of individuals through genetic operation such as reproduction,crossover,mutation,etc.The experimental results show that this algorithm is superior to genetic algorithm and weighted least connection algorithm.
An efficient data flow pattern and corresponding hardware architecture for efficient three-step search(E3SS) algorithm are preseated for motion estimation.The data flow exploits the overlap of reference data among the search points to reduce data memory accesses that are the most power consuming operations in motion estimation.The corresponding hardware architecture implements the search for two different patterns of E3SS efficiently.Simulation results show that the average number of cycles required to make a block matching for different test sequences is 280,which fulfills the speed requirement of MPEG-4 and HDTV standards.