extrsubs.js

const Jimp = require('jimp')
const { readdirSync } = require('fs')
const { join } = require('path')

const floodfill = require('./floodfill')

const sq = i => Math.pow(i, 2)
const distsq = ([ r1, g1, b1 ], [ r2, g2, b2 ]) => sq(r1 - r2) + sq(g1 - g2) + sq(b1 - b2)

//// Variables

const t = sq(0.45 * 255) // difference threshold (distance to raw image)
const t2 = sq(0.35 * 255) // pixel selection threshold (distance to white/black)
const minPixels = 100 // min subtitle pixel amount to trigger file save

////

async function one(raw, sub, to) {

    // Read images
    const rawi = await Jimp.read(raw)
    const subi = await Jimp.read(sub)

    const r = rawi.bitmap.data
    const s = subi.bitmap.data

    // Select the mode (white if image is quite dark and black if image is quite luminous)
    let bsum = 0
    let wsum = 0
    rawi.scan(0, 0, rawi.bitmap.width, rawi.bitmap.height, function (x, y, idx) {
        const rColor = [ r[ idx + 0 ], r[ idx + 1 ], r[ idx + 2 ] ]

        bsum += distsq(rColor, [ 0, 0, 0 ])
        wsum += distsq(rColor, [ 255, 255, 255 ])
    })

    const usingWhite = bsum < wsum;

    console.log(usingWhite ? 'using white mode' : 'using black mode')
    
    // The result image
    const result = new Jimp(1280, 720, 0x000000FF)
    const res = result.bitmap.data

    // Number of triggered pixels
    let n = 0

    result.scan(0, 0, result.bitmap.width, result.bitmap.height, function (x, y, idx) {

        const sColor = [ s[ idx + 0 ], s[ idx + 1 ], s[ idx + 2 ] ]
        const rColor = [ r[ idx + 0 ], r[ idx + 1 ], r[ idx + 2 ] ]

        if (
            ( // If pixel in raw image and pixel in sub image are similar
                distsq(sColor, rColor) > t
                && ( // Filter white only if white mode or black only if black mode
                    (usingWhite && distsq(sColor, [ 255, 255, 255 ]) < t2)
                    || (!usingWhite && distsq(sColor, [ 0, 0, 0 ]) < t2)
                )
            )
        ) { // Flip the pixel in the result image
            res[ idx + 0 ] = 255
            res[ idx + 1 ] = 255
            res[ idx + 2 ] = 255
            n++
        }

    })

    console.log('Interesting pixel count:', n)

    // If image is interesting (likely to contain subtitle)
    if (n > minPixels) {
        
        // Flood fill if black mode (this helps tesseract-ocr read)
        if (!usingWhite) {
            //floodfill(result, 0, 0, [ 0, 0, 0 ])
        }

        // Write
        await result.write(to)

        return true
    }

    return false
}

const offset = 115

const filename = i => `${i.toString().padStart(7, '0')}.png`

;(async () => {

    for (let i = 342; true; i++) {
        const rawimg = join('..', 'raw', filename(i))
        const subimg = join('..', 'source', filename(i + offset))
        const resimg = join('..', 'output', filename(i))

        const b = await one(rawimg, subimg, resimg)
        console.log(rawimg + ' + ' + subimg + ' => ' + resimg + (b ? ': yesssss!' : ': nop.'))
    }

    console.log('alldone.')

})().catch(console.error)