openscad/BOSL2
1
0
Fork 0
mirror of https://github.com/revarbat/BOSL2.git synced 2025-08-13 20:44:20 +02:00
Code Issues Projects Releases Wiki Activity

Added edge detection to img2scad.html

Browse Source
This commit is contained in:
Alex Matulich
2025-05-20 21:06:56 -07:00
parent d92dfbc400
commit bba8874f67
1 changed files with 107 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files

157
scripts/img2scad.html
View File

@@ -7,8 +7,9 @@ Versions 1-5: 22 April 2025 - by Alex Matulich (collaborating with ChatGPT for c
Version 6: 23 April 2025 - added cropping UI
Version 7: 25 April 2025 - added contrast and threshold sliders
Version 8: 26 April 2025 - added file size estimate to output section
Version 9: 20 May 2025 - improved appearance UI, added Sobel edge detection
-->
<title>Image to OpenSCAD array, v8</title>
<title>Image to OpenSCAD array, v9</title><!-- REMEMBER TO CHANGE VERSION -->
<meta charset="UTF-8">
<style>
body { font-family: sans-serif; padding-left:1em; padding-right:1em;}
@@ -212,7 +213,7 @@ Alpha channel is ignored. After processing the image as desired, you may save it
<fieldset>
<legend>Appearance</legend>
<div style="float:right; border:1px solid green; padding:4px; text-align:center; font-size:smaller;">See:<br><a href="https://en.wikipedia.org/wiki/Luma_(video)" target="_blank">Luma</a></div>
<div style="float:right; border:1px solid green; padding:4px; text-align:center; font-size:smaller;">See:<br><a href="https://en.wikipedia.org/wiki/Luma_(video)" target="_blank">Luma</a></div>
<input type="radio" name="grayModel" value="ntsc" checked><label for "grayModel" class="tooltip"> NTSC grayscale formula
<span class="tooltiptext">NTSC Y&prime; = 0.299R + 0.587G + 0.114B<br>Rec. 601: Average human perception of color luminance</span></label><br>
<input type="radio" name="grayModel" value="linear"><label for="grayModel" class="tooltip"> sRGB linear luminance
@@ -223,7 +224,10 @@ Alpha channel is ignored. After processing the image as desired, you may save it
</div>
<div style="margin:8px 0;">
<label for="blurRadius">Gaussian blur radius (pixels):</label>
<input type="number" id="blurRadius" size="5" min="0" max="20" value="0">
<input type="number" id="blurRadius" size="5" min="0" max="20" value="0"><br>
<label for="sobelRadius" class="tooltip">Edge detect radius (pixels):
<span class="tooltiptext">Sobel filter uses own radius if Gaussian blur=0</span></label>
<input type="number" id="sobelRadius" size="5" min="0" max="20" value="0">
</div>
<div class="slider-row">
@@ -287,6 +291,7 @@ Alpha channel is ignored. After processing the image as desired, you may save it
const cropRight = document.getElementById('cropRight');
const cropBottom = document.getElementById('cropBottom');
const blurRadiusInput = document.getElementById('blurRadius');
const sobelRadiusInput = document.getElementById('sobelRadius');
const contrastInput = document.getElementById('contrast');
const contrastValue = document.getElementById('contrastValue');
const thresholdInput = document.getElementById('threshold');
@@ -327,66 +332,113 @@ Alpha channel is ignored. After processing the image as desired, you may save it
// image processing functions
function applyGaussianBlur(matrix, radius) {
if (radius <= 0) return matrix;
function gaussianKernel1D(radius) {
const sigma = radius > 0 ? radius / 3 : 1;
const kernel = [];
let sum = 0;
for (let i = -radius; i <= radius; i++) { // kernel size = 2 * radius + 1;
const value = Math.exp(-(i * i) / (2 * sigma * sigma));
for (let i = -radius; i <= radius; i++) {
const value = Math.exp(- (i * i) / (2 * sigma * sigma));
kernel.push(value);
sum += value;
}
kernel.forEach((v, i) => kernel[i] = v / sum);
return kernel.map(v => v / sum);
}
function sobelDerivativeKernel(size) {
const half = Math.floor(size / 2);
const kernel = [];
for (let i = -half; i <= half; i++) {
kernel.push(i);
}
const norm = kernel.reduce((acc, val) => acc + Math.abs(val), 0) || 1;
return kernel.map(v => v / norm);
}
function convolve1DHorizontal(matrix, kernel, normalize=true) {
const width = matrix[0].length;
const height = matrix.length;
const horizontalBlur = [];
// blur pixels horizontally, put in horizontalBlur[]
const r = Math.floor(kernel.length / 2);
const result = [];
for (let y = 0; y < height; y++) {
horizontalBlur[y] = [];
result[y] = [];
for (let x = 0; x < width; x++) {
let val = 0;
let sum = 0;
let weightSum = 0;
for (let k = -radius; k <= radius; k++) {
for (let k = -r; k <= r; k++) {
const nx = x + k;
if (nx >= 0 && nx < width) {
val += matrix[y][nx] * kernel[k + radius];
weightSum += kernel[k + radius];
sum += matrix[y][nx] * kernel[k+r];
weightSum += kernel[k+r];
}
}
horizontalBlur[y][x] = val / weightSum;
result[y][x] = normalize ? (weightSum !== 0 ? sum / weightSum : 0) : sum;
}
}
// blur pixels vertically in horizontalBlur[], return result in output[]
const output = [];
return result;
}
function convolve1DVertical(matrix, kernel, normalize=true) {
const width = matrix[0].length;
const height = matrix.length;
const r = Math.floor(kernel.length / 2);
const result = [];
for (let y = 0; y < height; y++) {
output[y] = [];
result[y] = [];
for (let x = 0; x < width; x++) {
let val = 0;
let sum = 0;
let weightSum = 0;
for (let k = -radius; k <= radius; k++) {
for (let k = -r; k <= r; k++) {
const ny = y + k;
if (ny >= 0 && ny < height) {
val += horizontalBlur[ny][x] * kernel[k + radius];
weightSum += kernel[k + radius];
sum += matrix[ny][x] * kernel[k+r];
weightSum += kernel[k+r];
}
}
output[y][x] = val / weightSum;
result[y][x] = normalize ? (weightSum !== 0 ? sum / weightSum : 0) : sum;
}
}
return output;
return result;
}
function computeEdgeMagnitude(gx, gy) {
const height = gx.length;
const width = gx[0].length;
const result = [];
for (let y = 0; y < height; y++) {
result[y] = [];
for (let x = 0; x < width; x++) {
const mag = Math.sqrt(gx[y][x] ** 2 + gy[y][x] ** 2);
result[y][x] = mag;
}
}
return result;
}
function applyGaussianBlur(matrix, blurRadius) {
const gKernel = gaussianKernel1D(blurRadius)
g1 = convolve1DVertical(matrix, gKernel);
return convolve1DHorizontal(g1, gKernel);
}
function applySobel(matrix, sobelRadius, blurRadius) {
if (sobelRadius <= 0) return matrix; // No edge detection
const sobelSize = 2 * sobelRadius + 1;
const dKernel = sobelDerivativeKernel(sobelSize);
let gblur = blurRadius === 0 ? applyGaussianBlur(matrix, sobelRadius) : matrix;
gx = convolve1DHorizontal(gblur, dKernel, false);
gy = convolve1DVertical(gblur, dKernel, false);
return computeEdgeMagnitude(gx, gy);
}
function sigmoid(z) { return 1.0 / (1+Math.exp(-z)); } // used by contrastAdj
function contrastAdj(brightness) { // return an adjusted brightness based on contrast and threshold
const x = brightness/255.0;
const c = 2.0*contrast; // attempt to balance the sigmoid response to the contrast control
const sigterm = sigmoid(-c*threshold);
const adj = contrast>100.0 ? (x<threshold ? 0 : x>threshold ? 1 : threshold) // jump to 100% contrast at max contrast
: (sigmoid(c*(x-threshold)) - sigterm) / (sigmoid(c*(1.0-threshold)) - sigterm);
return adj * 255.0;
const x = brightness/255.0;
const c = 2.0*contrast; // attempt to balance the sigmoid response to the contrast control
const sigterm = sigmoid(-c*threshold);
const adj = contrast>100.0 ? (x<threshold ? 0 : x>threshold ? 1 : threshold) // jump to 100% contrast at max contrast
: (sigmoid(c*(x-threshold)) - sigterm) / (sigmoid(c*(1.0-threshold)) - sigterm);
return adj * 255.0;
}
function processImage() {
@@ -443,7 +495,11 @@ Alpha channel is ignored. After processing the image as desired, you may save it
const blurRadius = parseInt(blurRadiusInput.value) || 0;
const blurredMatrix = applyGaussianBlur(brightnessMatrix, blurRadius);
// crop the blurred matrix, gather min and max values in crop area
// apply Sobel edge detection
const sobelRadius = parseInt(sobelRadiusInput.value) || 0;
const sobelMatrix = applySobel(blurredMatrix, sobelRadius, blurRadius);
// crop the matrix, gather min and max values in crop area
const cropMatrix = [];
let cropx1 = parseInt(cropID[edgeID[2]].value) || 0;
let cropx2 = parseInt(cropID[edgeID[0]].value) || 0;
@@ -454,9 +510,9 @@ Alpha channel is ignored. After processing the image as desired, you may save it
for (let y=cropy1; y<uncropDim.height-cropy2; y++) {
const row = [];
for(let x=cropx1; x<uncropDim.width-cropx2; x++) {
row.push(blurredMatrix[y][x]);
min = Math.min(min, blurredMatrix[y][x]);
max = Math.max(max, blurredMatrix[y][x]);
row.push(sobelMatrix[y][x]);
min = Math.min(min, sobelMatrix[y][x]);
max = Math.max(max, sobelMatrix[y][x]);
}
cropMatrix.push(row);
}
@@ -524,6 +580,7 @@ Alpha channel is ignored. After processing the image as desired, you may save it
flipV = flipH = false;
resizeWidthInput.value = "100";
blurRadiusInput.value = "0";
sobelRadiusInput.value = "0";
invertBrightnessCheckbox.checked = invertBrightness = false;
contrastInput.value = contrastValue.textContent = "0";
contrast = 0.0001;
@@ -568,7 +625,7 @@ Alpha channel is ignored. After processing the image as desired, you may save it
// set up event listeners for all the input gadgets
[blurRadiusInput, contrastInput, thresholdInput,
[blurRadiusInput, sobelRadiusInput, contrastInput, thresholdInput,
...document.querySelectorAll('input[name="grayModel"]')].forEach(el => el.addEventListener('input', processImage));
resizeWidthInput.addEventListener('input', function () {
@@ -680,21 +737,21 @@ Alpha channel is ignored. After processing the image as desired, you may save it
processImage();
});
const Gbyte = 1073741824.0;
const Mbyte = 1048576.0;
const Kbyte = 1024.0;
// update file size estimate based on normalize type and size of output image
const Gbyte = 1073741824.0;
const Mbyte = 1048576.0;
const Kbyte = 1024.0;
// update file size estimate based on normalize type and size of output image
function updateKbytes() {
// length of a number for [0,1] range: mostly 6 characters "0.xxx," but occasionally less, using 5.95.
// length of a number for [0,255] range: assume 0-255 are uniformly distributed, use weighted average of digits plus comma
// length of a number for [0,1] range: mostly 6 characters "0.xxx," but occasionally less, using 5.95.
// length of a number for [0,255] range: assume 0-255 are uniformly distributed, use weighted average of digits plus comma
const avglen = normalizeToUnitCheckbox.checked ? 5.95 : (10.0+90.0*2.0+156.0*3.0)/256.0+1.0;
// each row has 6 extra characters " [],\r\n" at most, plus 5 characters after array name and 4 characters at the end
// each row has 6 extra characters " [],\r\n" at most, plus 5 characters after array name and 4 characters at the end
const estsize = (avglen*cropDim.width + 6.0) * cropDim.height + 9 + arrayName.value.length;
let unitName = "bytes";
let unit = 1.0;
if (estsize > Gbyte) { unit = Gbyte; unitName = "GiB"; }
else if (estsize > Mbyte) { unit = Mbyte; unitName = "MiB"; }
else if (estsize > 10.0*Kbyte) { unit = Kbyte; unitName = "KiB"; }
let unitName = "bytes";
let unit = 1.0;
if (estsize > Gbyte) { unit = Gbyte; unitName = "GiB"; }
else if (estsize > Mbyte) { unit = Mbyte; unitName = "MiB"; }
else if (estsize > 10.0*Kbyte) { unit = Kbyte; unitName = "KiB"; }
const sizeOut = (estsize/unit).toFixed(unit==1.0?0:1);
kbytes.textContent = `${sizeOut} ${unitName}`;
}
@@ -714,8 +771,8 @@ Alpha channel is ignored. After processing the image as desired, you may save it
const arrayContent = grayscaleMatrix.map(row => {
return " [" + row.map(val => useUnit ? parseFloat((val/255.0).toFixed(3)) : val).join(",") + "]";
}).join(",\n");
const introcomment = " = [ // " + cropDim.width + "×" + cropDim.height + "\n";
const dimSuffix = "_"+cropDim.width + "x" + cropDim.height
const introcomment = " = [ // " + cropDim.width + "×" + cropDim.height + "\n";
const dimSuffix = "_"+cropDim.width + "x" + cropDim.height
const openscadArray = (arrayName.value.length>0 ? arrayName.value : 'image_array') + introcomment + arrayContent + "\n];";
const blob = new Blob([openscadArray], { type: "text/plain" });
let filename = (arrayName.value.length>0 ? arrayName.value : "image_array") + dimSuffix + '.scad';