8000 miniPaint/libs/sift.js at v3.2 · P3PO/miniPaint · GitHub
[go: up one dir, main page]

Skip to content

Latest commit

 

History

History
183 lines (169 loc) · 6.37 KB

sift.js

File metadata and controls

183 lines (169 loc) · 6.37 KB
/**
* SIFT: scale-invariant-feature-transform, keypoints
*
* @author ViliusL
*/
/* global HELPER, IMAGE, ImageFilters, LAYER, canvas_active */
var SIFT = new SIFT_CLASS();
function SIFT_CLASS() {
/**
* contrast check, smaller - more points, better accuracy, but slower
*/
var avg_offset = 50;
/**
* how much pixels to check for each side to get average
*/
var avg_step = 4;
//generate key points for image
this.generate_keypoints = function (canvas, show_points) {
var W = canvas.width;
var H = canvas.height;
HELPER.timer_init();
//check whitespace
var trim_details = IMAGE.trim_info(canvas);
W = W - trim_details.left - trim_details.right;
H = H - trim_details.top - trim_details.bottom;
//make copy
var clone = document.createElement("canvas");
clone.width = W;
clone.height = H;
var ctx = clone.getContext("2d");
ctx.drawImage(canvas, -trim_details.left, -trim_details.top, canvas.width, canvas.height);
//greyscale
var imageData = ctx.getImageData(0, 0, W, H);
var filtered = ImageFilters.GrayScale(imageData); //add effect
ctx.putImageData(filtered, 0, 0);
//make few copies and blur each
var n = 5;
var copies = [];
for (var i = 0; i < n; i++) {
var tmp_canvas = document.createElement("canvas");
tmp_canvas.width = W;
tmp_canvas.height = H;
var ctx_i = tmp_canvas.getContext("2d");
ctx_i.drawImage(clone, 0, 0);
//gausian blur
var imageData = ctx_i.getImageData(0, 0, W, H);
var filtered = ImageFilters.GaussianBlur(imageData, i + 0.5); //add effect
ctx_i.putImageData(filtered, 0, 0);
copies.push(tmp_canvas);
}
//find extreme points
var points = [];
var n0 = avg_step * 2 + 1;
for (var c = 1; c < copies.length - 1; c++) {
var imageData = copies[c].getContext("2d").getImageData(0, 0, W, H).data;
var imageData0 = copies[c - 1].getContext("2d").getImageData(0, 0, W, H).data;
var imageData2 = copies[c + 1].getContext("2d").getImageData(0, 0, W, H).data;
for (var j = avg_step; j < H - avg_step; j++) {
for (var i = avg_step; i < W - avg_step; i++) {
var x = (i + j * W) * 4;
if (imageData[x + 3] == 0)
continue; //transparent
if (imageData[x] < imageData[x - 4] || imageData[x] < imageData[x + 4] || imageData[x] > imageData[x - 4] || imageData[x] > imageData[x + 4]) {
var x_pre = (i + (j - 1) * W) * 4;
var x_post = (i + (j + 1) * W) * 4;
//calc average
var area_average = 0;
for (var l = -avg_step; l <= avg_step; l++) {
var avgi = (i + (j - l) * W) * 4;
for (var a = -avg_step; a <= avg_step; a++) {
area_average += imageData[avgi + 4 * a];
}
}
area_average = area_average / (n0 * n0);
//max
if (imageData[x] + avg_offset < area_average) {
var min = Math.min(imageData[x_pre - 4], imageData[x_pre], imageData[x_pre + 4], imageData[x - 4], imageData[x + 4], imageData[x_post - 4], imageData[x_post], imageData[x_post + 4]);
if (imageData[x] <= min) {
var min0 = Math.min(imageData0[x_pre - 4], imageData0[x_pre], imageData0[x_pre + 4], imageData0[x - 4], imageData0[x + 4], imageData0[x_post - 4], imageData0[x_post], imageData0[x_post + 4]);
if (imageData[x] <= min0) {
var min2 = Math.min(imageData2[x_pre - 4], imageData2[x_pre], imageData2[x_pre + 4], imageData2[x - 4], imageData2[x + 4], imageData2[x_post - 4], imageData2[x_post], imageData2[x_post + 4]);
if (imageData[x] <= min2)
points.push({
x: i + trim_details.left,
y: j + trim_details.top,
w: Math.round(area_average - imageData[x] - avg_offset)
});
}
}
continue;
}
//min
if (imageData[x] - avg_offset > area_average) {
var max = Math.max(imageData[x_pre - 4], imageData[x_pre], imageData[x_pre + 4], imageData[x - 4], imageData[x + 4], imageData[x_post - 4], imageData[x_post], imageData[x_post + 4]);
if (imageData[x] >= max) {
var max0 = Math.max(imageData0[x_pre - 4], imageData0[x_pre], imageData0[x_pre + 4], imageData0[x - 4], imageData0[x + 4], imageData0[x_post - 4], imageData0[x_post], imageData0[x_post + 4]);
if (imageData[x] >= max0) {
var max2 = Math.max(imageData2[x_pre - 4], imageData2[x_pre], imageData2[x_pre + 4], imageData2[x - 4], imageData2[x + 4], imageData2[x_post - 4], imageData2[x_post], imageData2[x_post + 4]);
if (imageData[x] >= max2){
points.push({
x: i + trim_details.left,
y: j + trim_details.top,
w: Math.round(imageData[x] - area_average - avg_offset)
});
}
}
}
}
}
}
}
}
//make unique
for (var i = 0; i < points.length; i++) {
for (var j = 0; j < points.length; j++) {
if (i != j && points[i].x == points[j].x && points[i].y == points[j].y) {
points.splice(i, 1);
i--;
break;
}
}
}
//show points?
if (show_points === true) {
var time = HELPER.timer('', true);
console.log('key points: ' + points.length + ", " + time);
LAYER.layer_add();
var size = 3;
canvas_active().fillStyle = "#ff0000";
for (var i in points) {
var point = points[i];
canvas_active().beginPath();
canvas_active().rect(point.x - Math.floor(size / 2) + 1, point.y - Math.floor(size / 2) + 1, size, size);
canvas_active().fill();
}
}
else {
//sort by weights
points.sort(function (a, b) {
return parseFloat(b.w) - parseFloat(a.w);
});
return {
points: points,
trim_details: trim_details
};
}
};
//returns average value of requested area from greyscale image
//area = {x, y, w, h}
this.get_area_average = function (area, imageData, i, j, size) {
var imgData = imageData.data;
var sum = 0;
var n = 0;
size = size / 100; //prepare to use 1-100% values
var stop_x = i + Math.round(size * area.x) + Math.round(size * area.w);
var stop_y = j + Math.round(size * area.y) + Math.round(size * area.h);
var img_width4 = imageData.width * 4;
var k0, k;
for (var y = j + Math.round(size * area.y); y < stop_y; y++) {
k0 = y * img_width4;
for (var x = i + Math.round(size * area.x); x < stop_x; x++) {
k = k0 + (x * 4);
sum = sum + imgData[k];
n++;
}
}
return Math.round(sum / n);
};
}
2BDA
0