We propose an approach to produce high quality, visually cognitive QR codes, which we call QArt – a new generation of QR codes that are machine-readable and visually perceivable by humans simultaneously. First, a pattern readability function is constructed wherein a probability distribution is learned which identifies replaceability of modules which are the collections of data inside QR code. Then, given a text tag, an input image is taken and based on the Grayscale label on that image, a high quality, visual appearing QR code is generated. A user just needs a phone with picture taking capability to extract the encoded data from these QR codes which may include URL, music, images, or a plethora of other digital content. Some advertisers use QR codes to promote their web presence since the smartphones can read and immediately access the URL encoded in it. As an instance of application of the technique proposed in this paper, such advertisers may generate a QR code on their company logo which contains the encoded URL, for advertising or promoting purpose.