An image-taking apparatus is disclosed which achieves focusing in a reduce time with higher accuracy of an in-focus position search in a TV-AF method to minimize occurrence of unnatural focus changes. The image-taking apparatus includes an image-taking element, a first detector which outputs a focus evaluation value signal based on a predetermined frequency component of an output signal of the image-pickup element, a second detector which outputs a detection signal different from the focus evaluation value signal, and a controller which performs first processing of obtaining information for an in-focus position search of the focus lens based on the focus evaluation value signal. The controller performs second processing of obtaining information on an in-focus position based on the detection signal from the second detector, and performs third processing different depending on a comparison result between the information from the first processing and the information obtained from the second processing.