There is no conclusive explanation of why ~80 per cent of planetary nebulae (PNe) are non-spherical. In the Binary Hypothesis, a binary interaction is a preferred channel to form a non-spherical PN. A fundamental step to corroborate or disprove the Binary Hypothesis is to estimate the binary fraction of central stars of PNe (CSPNe) and compare it with a prediction based on the binary fraction of the progenitor, main-sequence population. In this paper, the second in a series, we search for spatially unresolved I- and J-band flux excess in an extended sample of 34 CSPN by a refined measurement technique with a better quantification of the uncertainties. The detection rate of I- (J-)band flux excess is 32 ± 16 per cent (50 ± 24 per cent). This result is very close to what was obtained in Paper I with a smaller sample. We account conservatively for unobserved cool companions down to brown dwarf luminosities, increasing these fractions to 40 ± 20 per cent (62 ± 30 per cent). This step is very sensitive to the adopted brightness limit of our survey. Accounting for visual companions increases the binary fraction to 46 ± 23 per cent (71 ± 34 per cent). These figures are lower than in Paper I. The error bars are better quantified, but still unacceptably large. Taken at face value, the current CSPN binary fraction is in line with the main-sequence progenitor population binary fraction. However, including white dwarfs companions could increase this fraction by as much as 13 (21) per cent points.