Homec4science
Diffusion ShoulderDB_ImportSegmentMeasureAnalyze (aterrier-plotUpdate) Browse

R8218/CT/private/dicom_encode_rle.m
e412f5003517aterrier-plotUpdate

Last Change
Raw File

dicom_encode_rle.m
View Options

function [fragments, frames] = dicom_encode_rle(pixel_cells, bits_allocated, dims)
%DICOM_ENCODE_RLE Run-length encode pixel cells.
% [FRAGMENTS, LIST] = DICOM_ENCODE_RLE(PIXCELLS, BITS, DIMS) compresses
% the pixel cells PIXCELLS using run-length encoding. BITS is the
% number of bits allocated for each pixel cell, and DIMS is the
% original dimensions of the image stored in PIXCELLS, which is now a
% vector. FRAGMENTS is a cell array containing the encoded frames (as
% UINT8 data) from the compressor. LIST is a vector of indices to the
% first fragment of each compressed frame of a multiframe image.
%
% See also DICOM_ENCODE_JPEG_LOSSY.
% Copyright 1993-2011 The MathWorks, Inc.
% Run-length encoded pixel cells have the following form:
%
% RLE Header (16 UINT32 components)
% - The first component is number of RLE segments (0 - 15)
% - The remaining 15 components are offsets to the beginning of each
% segment (or 0 if there are no more segments).
% RLE Segment 1
% RLE Segment 2
% etc.
%
% The segments are generated by stripping off successive bytes from the
% composite pixel code, starting with the most significant byte [1].
% The stripped bytes are passed through the RLE coder to create the
% compressed segment.
%
% Individual image rows must be sent through the coder separately even if
% they are part of the same segment. This prevents a run from spanning
% multiple rows. The compressed rows are concatenated within each
% segment.
%
% If a composite pixel cell does not end on byte boundaries, 0 bits are
% appended to the least significant byte. This will happen if the number
% of bits allocated for pixels and overlays (0028,0100) is not divisible
% by 8.
%
% Sequences of bytes can be compressed into an arbitrary number of
% fragments. This allows for encoding the image into a set of "strips".
% The only restriction is that each fragment must contain the same number
% of pixel cell elements in each segment (e.g., the same number of red,
% green, and blue samples, etc.).
%
% [1] - A composite pixel code is one pixel cell in the sequence of pixel
% cells. The unstated conclusion of using composite pixel codes is
% that planar configuration (0028,0006) must be 0 if more than one
% sample is present and RLE coding is used.
%
% See Annex G in part 5 of the DICOM spec for more details.
% DICOM_ENCODE_PIXEL_CELLS explicitly creates pixel cells that are byte
% aligned.
if (rem(bits_allocated, 8) ~= 0)
error(message('images:dicom_encode_rle:badBitsAllocated'))
end
% Reshape data to make segment creation easy.
if (length(dims) >= 3)
num_segments = (bits_allocated / 8) * dims(3);
else
num_segments = (bits_allocated / 8);
end
if (num_segments > 15)
error(message('images:dicom_encode_rle:tooManySegments'))
end
if (numel(dims) == 4)
numFrames = dims(4);
else
numFrames = 1;
end
swap = determine_swap;
pixel_cells = dicom_typecast(pixel_cells, 'uint8', swap);
pixel_cells = reshape(pixel_cells, [num_segments, ...
(numel(pixel_cells) / num_segments)]);
offset = 0;
fragments = cell(1, numFrames);
for f = 1:numFrames
encoded_data = [];
% Encode each segment.
segment_lengths = zeros(1, num_segments);
for p = 1:num_segments
segment = [];
% Encode each row.
for q = 1:dims(1)
start_idx = offset + (q - 1) * dims(2) + 1;
end_idx = offset + q * dims(2);
encoded_row = dicom_encode_rle_segment(pixel_cells(p, start_idx:end_idx));
segment = [segment; encoded_row]; %#ok<AGROW>
end
% Store segment.
encoded_data = [encoded_data; segment]; %#ok<AGROW>
segment_lengths(p) = numel(segment);
end
offset = end_idx;
% Create RLE header.
header = repmat(uint32(0), [16 1]);
header(1) = uint32(num_segments);
header(2) = uint32(64);
for p = 2:(num_segments)
header(p + 1) = uint32(64 + sum(segment_lengths(1:(p - 1))));
end
encoded_data = [dicom_typecast(header, 'uint8', swap); encoded_data]; %#ok<AGROW>
% Create output from fragments.
fragments{f} = encoded_data;
end
frames = 1:numFrames;
% This is essentially how dicom_rle_encode_segment() works.
% function coded_bytes = rle_coder(X)
%
% X = double(X(:));
%
% current_pos = 1;
% max_pos = numel(X);
%
% % Find runs. Where the diff is 0, there is a repeated value. The first run
% % is not included.
% X_diff = diff(X);
% X_reps = find(X_diff == 0);
% rep_pix = [X_reps(1); X_reps(find(diff(X_reps) > 1) + 1); (max_pos + 1)];
%
% all_runs = find(X_diff ~= 0) + 1;
%
% rep_number = 1;
% next_rep = rep_pix(rep_number);
%
% coded_values = [];
% more_data = 1;
%
% while (more_data)
%
% if (current_pos ~= next_rep)
%
% % Handle literal values between runs.
%
% % Only 128 literal values can be encoded in one literal run.
% while ((next_rep - current_pos) > 128)
%
% coded_values = [coded_values
% 127
% X(current_pos:(current_pos + 127))];
% current_pos = current_pos + 128;
%
% end
%
% run_length = next_rep - current_pos;
% coded_values = [coded_values
% (run_length - 1)
% X(current_pos:(next_rep - 1))];
%
% % All literal values until the next repetitive run are encoded.
% current_pos = next_rep;
%
% else
%
% % Handle repetitive runs.
%
% % Find next run of different values.
% idx = find(all_runs == current_pos);
%
% if (isempty(idx))
%
% % This run appeared at the very beginning of the data.
% if (isempty(all_runs))
%
% % There is only one run.
% next_run = max_pos + 1;
%
% else
%
% next_run = all_runs(1);
%
% end
%
% elseif (idx == numel(all_runs))
%
% % This is the last run.
% next_run = max_pos + 1;
%
% else
%
% next_run = all_runs(idx + 1);
%
% end
%
% while ((next_run - current_pos) > 128)
%
% coded_values = [coded_values
% -127
% X(current_pos)];
% current_pos = current_pos + 128;
%
% end
%
% run_length = next_run - current_pos;
% coded_values = [coded_values
% (-run_length + 1)
% X(current_pos)];
%
% current_pos = current_pos + run_length;
%
% rep_number = rep_number + 1;
% next_rep = rep_pix(rep_number);
%
% end
%
% if (current_pos > max_pos)
%
% more_data = 0;
%
% end
%
% end
%
% % Convert double data to bytestream.
% %
% % The next statement maps the signed values in the range [-128, 255] to
% % [0, 255], the valid range for UINT8. Values in the range [0, 255] are
% % unchanged; values in [-128, -1] are mapped to [128, 255].
% %
% % Converting negative values is okay because the bit pattern for a
% % converted value as a UINT8 is the same as the negative value's bit
% % pattern as an INT8. The RLE decoder algorithm will read the next byte
% % as a signed or unsigned value based on its own logic; it suffices for
% % this function to write indistinguishable bytes (signed or unsigned).
% coded_bytes = uint8(rem((coded_values + 256), 256));
function swap = determine_swap
% Determine if endianness of data needs swapped.
[~, ~, mach] = fopen(1);
% RLE compressed data must be written to file as IEEE-LE.
if (isempty(strfind(lower(mach), 'ieee-le')))
swap = 1;
else
swap = 0;
end
c4science · Help