The probability depends on how you would work it out... 

On one hand, the odds of a picture you looking at being of any given celebrity are pretty slim (say for example, 1/10,000) and of hearing any given celebrity voice are also pretty slim (eg 1/10,000) so the odds of coming across both at once are close to the product of these: 1/100,000,000.  Pretty unlikely. 

But the odds are weighted.  The odds of finding a picture of Mr Fry in any given publication might be 1/10,000, but Stephen Fry is a brainy celebrity and you were reading a brainy magazine.  If Stephen Fry is going to feature in a magazine, it's a good bet that his picture will be there too, to draw the reader to the article.  Even in a magazine full of brainy types, such as New Scientist, the odds of looking at a picture which is of  Stephen Fry are pretty high, 1/100 for example. 

Then there's the radio.  You read New Scientist, so it's not a big coincidence that you were listening to a 'brainy' radio station (maybe you weren't, but I'm assuming it's a 'brainy' one if Stephen Fry is on it).  Again, the chances of Stephen Fry being on a brainy media outlet are actually pretty high.  If we assume they are also 1/100, you get a total odds of 1/10,000.  However, there is much more to consider.  The odds of hearing Stephen Fry on the radio go up if he's advertising a new show, and the odds of him advertising this show at peak hours goes up even further. (Advertising a new show also increases the likelihood of him, and therefore his picture, featuring in a magazine such as NS).

Really, you would need to know the total number of seconds he was on air for during a given period, and how long you listened for.  The resulting odds would in part depend on what given period you chose - an hour, a day, a week?  The choice would be arbitrary.  The question would then be something like "What are the chances that during the 35 minutes of Radio 2's breakfast show that I listened to on August 18, from 8.30 to 9.05, I would hear Stephen Fry's voice at exactly the same time as I looked at his picture in the latest issue of New Scientist?".  The probability largely comes down to timing because you are in the advertiser's target demographic.  You would also need to know how much time you spent looking at pictures of Stephen Fry in that period.  If you find his face fascinating and spend much of every day gazing at his images, the odds of hearing him while looking at his photo are increased.  (Not saying you do this, just illustrating a point ;) )

As you can see, how you would work out the odds depends on having quite a lot of information available and also the chosen parameters used.  It's worth learning about statistics like this, despite the tediousness of them, because they are easy to mis-represent in the media and we need to educate ourselves against inept or misleading statistics.  Note that a phrase like "what are the odds of looking at a picture of Stephen Fry" is different to "what are the odds of a picture you are looking at being of Stephen Fry".  In the former, you have to consider the odds of you actually looking at a picture.  In the latter, the odds of looking at a picture are 100%, because the statement tells you you are already looking at one.