你可以为每个拉丁字母(大写和小写)创建一个单独的画布进行比较。每当遇到不在Latin-1范围内的字符时,就为其创建一个新的画布,并使用
image diff算法将其与每个拉丁字母进行比较。用最接近的匹配替换非拉丁字符。
例如:
var latinize = (function () {
var latinLetters = [],
canvases = [],
size = 16,
halfSize = size >> 1;
function makeCanvas(chr) {
var canvas = document.createElement('canvas'),
context = canvas.getContext('2d');
canvas.width = size;
canvas.height = size;
context.textBaseline = 'middle';
context.textAlign = 'center';
context.font = (halfSize) + "px sans-serif";
context.fillText(chr, halfSize, halfSize);
return context;
}
function nextChar(chr) {
return String.fromCharCode(chr.charCodeAt(0) + 1);
}
function setupRange(from, to) {
for (var chr = from; chr <= to; chr = nextChar(chr)) {
latinLetters.push(chr);
canvases.push(makeCanvas(chr));
}
}
function calcDistance(ctxA, ctxB) {
var distance = 0,
dataA = ctxA.getImageData(0, 0, size, size).data,
dataB = ctxB.getImageData(0, 0, size, size).data;
for (var i = dataA.length; i--;) {
distance += Math.abs(dataA[i] - dataB[i]);
}
return distance;
}
setupRange('a', 'z');
setupRange('A', 'Z');
setupRange('', '');
return function (text) {
var result = "",
scores, canvas;
for (var i = 0; i < text.length; i++) {
if (text.charCodeAt(i) < 128) {
result += text.charAt(i);
continue;
}
scores = [];
canvas = makeCanvas(text.charAt(i));
for (var j = 0; j < canvases.length; j++) {
scores.push({
glyph: latinLetters[j],
score: calcDistance(canvas, canvases[j])
});
}
scores.sort(function (a, b) {
return a.score - b.score;
});
result += scores[0].glyph;
}
return result;
}
}());
这将把你的测试字符串翻译成“Nike Dunk High SB 'Ugly Sweater'现已上市”。另一种方法是创建一个巨大的数据结构,将所有类似字符映射到它们的Latin-1等效项,就像@willy答案中的库所做的那样。这对于“浏览器JavaScript”来说非常重,可能不适合发送给客户端,正如您从该项目的源代码中看到的那样。
http://jsfiddle.net/Ly5Lt/4/