Donald G. Bailey

A novel approach to real-time bilinear interpolation

Bilinear interpolation is often used to improve image quality after performing spatial transformation operations such as digital zooming or rotation. In the traditional case where the input coordinates appear in a raster-based fashion, the required pixel values can be obtained from the previous calculation, the frame buffer and a single line cache. This paper presents a novel approach to performing real-time bilinear interpolation that is useful in applications such as lens distortion correction where the input coordinates follow a curved path that spans multiple rows. To help retrieve the required pixels in a single clock cycle under imposed data bandwidth constraints a unique caching system has been devised. In the event that constraints make it impossible to obtain the four required pixel values, the approach performs a modified three-point interpolation. An example field programmable gate array implementation of the bilinear interpolation method used in conjunction with a lens distortion correction algorithm has been successfully completed.

Properties, implementations and applications of rank filters

Abstract The results of research on rank filters are presented. The relationship of rank filters with other filters is briefly discussed. The main properties of rank filters are listed and an explanation is given for these properties. Several software and hardware implementations of the filter are described. Major applications to image processing are discussed, including noise smoothing, cluster detection, skeletization, edge enhancement and edge detection.

FPGA implementation of a Single Pass Connected Components Algorithm

The classic connected components labelling algorithm requires two passes through an image. This paper presents an algorithm that allows the connected components to be analysed in a single pass by gathering data on the regions as they are built. This avoids the need for buffering the image, making it ideally suited for processing streamed images on an FPGA or other embedded system with limited memory. An FPGA-based implementation is described, emphasising the modifications made to the algorithm to enable it to satisfy timing constraints.

An efficient euclidean distance transform

Within image analysis the distance transform has many applications. The distance transform measures the distance of each object point from the nearest boundary. For ease of computation, a commonly used approximate algorithm is the chamfer distance transform. This paper presents an efficient linear- time algorithm for calculating the true Euclidean distance-squared of each point from the nearest boundary. It works by performing a 1D distance transform on each row of the image, and then combines the results in each column. It is shown that the Euclidean distance squared transform requires fewer computations than the commonly used 5x5 chamfer transform.

Test bed for number plate recognition applications

A flexible software based platform for number plate recognitions applications is described. It breaks the processing into several explicit modules, with the implementation for each module provided by a software plug-in using a DLL interface. The modular structure greatly facilitates the comparison between different algorithms for a particular step, and allows code to be reused between applications.

Optimised single pass connected components analysis

Classical connected components labelling algorithms are unsuitable for real-time processing of streamed images on an FPGA because they require two passes through the image. Recently, a single-pass algorithm was proposed that avoided the need to buffer an intermediate image. In this paper, a new single pass algorithm is described that is a considerable improvement over the existing algorithms. The new algorithm reassigns and reuses labels each row to minimise the size of both the equivalence and region data tables. The optimised single-pass algorithm reduces the worst case memory requirement by over 100 times that of the original algorithm (for measuring region area), and reduces the latency to only 1 row.

Connected components analysis of streamed images

Classical connected components labelling algorithms are unsuitable for real-time processing of streamed images on an FPGA because they require two passes through the image. The basic requirements of connected components analysis are investigated, and this led to a novel single-pass approach that avoids the need to buffer an intermediate image. This is further analysed to give an algorithm that is both memory efficient and has the minimum latency. The final algorithm reduces the memory by over 300 times, and reduces the latency by more than a factor of 2.

Range filters: Localintensity subrange filters and their properties

Abstract A local filter which uses the local-intensity subrange of pixel intensity values within a window is described. The range filter is an extension of the rank filter and has been found useful for detecting edges. The deterministic and noise properties of the range filter are described and compared with those of the commonly used Sobel filter.

Analysis of Errors in ToF Range Imaging With Dual-Frequency Modulation

Range imaging is a technology that utilizes an amplitude-modulated light source and gain-modulated image sensor to simultaneously produce distance and intensity data for all pixels of the sensor. The precision of such a system is, in part, dependent on the modulation frequency. There is typically a tradeoff between precision and maximum unambiguous range. Research has shown that, by taking two measurements at different modulation frequencies, the unambiguous range can be extended without compromising distance precision. In this paper, we present an efficient method for combining two distance measurements obtained using different modulation frequencies. The behavior of the method in the presence of noise has been investigated to determine the expected error rate. In addition, we make use of the signal amplitude to improve the precision of the combined distance measurement. Simulated results compare well to actual data obtained using a system based on the PMD19k range image sensor.

Super-resolution of bar codes

This article describes how under-resolved images of bar codes may be read by suitable processing. A two-dimensional im- age of a bar code with insufficient resolution to be able to resolve the individual bars is processed to give a high-resolution image. For this to work, the bar code (or camera) must be slightly rotated to give a fraction of a pixel offset between rows. Since the bars are straight, the offset relative to the first complete row of the bar code increases linearly with vertical position in the image. This offset be- tween rows results in a shift in phase that is proportional to both offset and frequency. A phase image is formed by Fourier trans- forming each row in the image, and retaining the phases. By sub- tracting the first row from subsequent rows of the phase image, a surface is fitted to give the offset between rows. A high-resolution image is then formed by interleaving the pixel values from rows where the offset is nearest to the new pixel spacing. This image appears blurred because of the area sampling caused by the sen- sor, combined with the low pass response of the camera electronics. By modeling the image capture system, the point spread function may be estimated and then removed by using inverse filtering in the frequency domain. The offset between the rows is then removed by using a linear phase filter. This allows the rows within the resultant image to be averaged to reduce noise.