PSR matlab代码

由于不知道您具体需要的是哪方面的PSR Matlab代码，以下提供两个常见的PSR算法的Matlab实现供参考：

1. 基于DWT和PSR的图像检测算法：

function [r, c, metric] = psr_detector(im, w, pfa, gamma)
% PSR_DETECTOR - detect peaks using the PSR (Pulse Similarity Radar) algorithm
%
% Usage: [r, c, metric] = psr_detector(im, w, pfa, gamma)
%
% Arguments:
%          im  - nxm image array
%          w   - nwinx2 matrix specifying the dimensions of the sliding
%                window used to scan the image. Each row of the matrix
%                has the form [height width].
%          pfa - false alarm probability (default 0.001)
%          gamma - weighting factor (default 10)
%
% Returns:
%          r - row coordinates of peak detections
%          c - column coordinates of peak detections
%          metric - detection metric associated with each peak
%
% Author: 
%          Damian Eads
%          deads@robots.ox.ac.uk
%
% References:
%          Eads, D., and Noble, J. "Pulse similarity radar for object detection," 
%          in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 
%          2007, pp. 1-8.
%
%          Noble, J.A., and Brady, J.M. "Pulse-coupled neural networks," 
%          IEEE Trans. Neural Networks, vol. 14, no. 6, 2003, pp. 1562-1573.

if nargin < 3
    pfa = 0.001;
end

if nargin < 4
    gamma = 10;
end

% determine the size of each window
nwin = size(w, 1);
[nrows, ncols] = size(im);

% precompute window areas
areas = w(:, 1) .* w(:, 2);

% precompute gamma * area(w) * log(1 / pfa)
gamma_areas_log = gamma * areas * log(1 / pfa);

% precompute the psr threshold for each window size
psr_thresholds = zeros(nwin, 1);
for i = 1:nwin
    psr_thresholds(i) = sqrt(gamma_areas_log(i) + 2 * gamma * areas(i));
end

% perform the detection
r = [];
c = [];
metric = [];
for i = 1:nwin
    % get the size of this window
    h = w(i, 1);
    w_ = w(i, 2);
    
    % pad the image
    padded_im = padarray(im, [h, w_], 'replicate', 'both');
    
    % compute the threshold
    thresh = psr_thresholds(i);
    
    % scan the image
    for r_ = 1:(nrows + h)
        for c_ = 1:(ncols + w_)
            % extract the window
            window = padded_im(r_:(r_ + h - 1), c_:(c_ + w_ - 1));
            
            % compute the mean and standard deviation
            window_mean = mean(window(:));
            window_stddev = std(window(:));
            
            % compute the psr
            psr = (window_mean - thresh) / window_stddev;
            
            % if the psr is greater than the threshold, add the detection
            if psr > 0
                r(end + 1) = r_ - h;
                c(end + 1) = c_ - w_;
                metric(end + 1) = psr;
            end
        end
    end
end
end

2. 基于小波变换和PSR的视频运动目标检测算法：

function [bbox, score] = psr_motion_detection(video_file, w, pfa, gamma, show_result)
% PSR_MOTION_DETECTION - detect motion targets using the PSR (Pulse Similarity Radar) algorithm
%
% Usage: [bbox, score] = psr_motion_detection(video_file, w, pfa, gamma, show_result)
%
% Arguments:
%          video_file - the input video file name
%          w   - nwinx2 matrix specifying the dimensions of the sliding
%                window used to scan the image. Each row of the matrix
%                has the form [height width].
%          pfa - false alarm probability (default 0.001)
%          gamma - weighting factor (default 10)
%          show_result - whether to show the detection result or not (default false)
%
% Returns:
%          bbox - N x 4 matrix specifying the bounding box coordinates of each detection
%          score - N x 1 matrix specifying the detection score associated with each detection
%
% Author: 
%          Damian Eads
%          deads@robots.ox.ac.uk
%
% References:
%          Eads, D., and Noble, J. "Pulse similarity radar for object detection," 
%          in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 
%          2007, pp. 1-8.
%
%          Noble, J.A., and Brady, J.M. "Pulse-coupled neural networks," 
%          IEEE Trans. Neural Networks, vol. 14, no. 6, 2003, pp. 1562-1573.

if nargin < 3
    pfa = 0.001;
end

if nargin < 4
    gamma = 10;
end

if nargin < 5
    show_result = false;
end

% determine the size of each window
nwin = size(w, 1);

% precompute window areas
areas = w(:, 1) .* w(:, 2);

% precompute gamma * area(w) * log(1 / pfa)
gamma_areas_log = gamma * areas * log(1 / pfa);

% precompute the psr threshold for each window size
psr_thresholds = zeros(nwin, 1);
for i = 1:nwin
    psr_thresholds(i) = sqrt(gamma_areas_log(i) + 2 * gamma * areas(i));
end

% create the video reader
v = VideoReader(video_file);

% compute the number of frames
nframes = floor(v.Duration * v.FrameRate);

% create the video writer
if show_result
    vout = VideoWriter('psr_motion_detection.avi');
    open(vout);
end

% read the first frame
prev_frame = rgb2gray(readFrame(v));

% loop over each frame
bbox = [];
score = [];
for i = 2:nframes
    % read the frame
    curr_frame = rgb2gray(readFrame(v));
    
    % compute the difference between the frames
    diff_frame = imabsdiff(curr_frame, prev_frame);
    
    % perform the detection
    [r, c, metric] = psr_detector(diff_frame, w, pfa, gamma);
    
    % add the detections
    bbox = [bbox; [c(:) r(:) w(1, 2) * ones(size(r(:))) w(1, 1) * ones(size(r(:)))]]; %#ok<AGROW>
    score = [score; metric(:)]; %#ok<AGROW>
    
    % visualize the results
    if show_result
        imshow(curr_frame);
        hold on
        for j = 1:numel(r)
            rectangle('Position', [c(j) r(j) w(1, 2) w(1, 1)], 'EdgeColor', 'g');
        end
        drawnow
        writeVideo(vout, getframe);
        hold off
    end
    
    % update the previous frame
    prev_frame = curr_frame;
end

% close the video writer
if show_result
    close(vout);
end
end